diff --git a/ai-at-work.mdx b/ai-at-work.mdx index eaa657e..d70598b 100644 --- a/ai-at-work.mdx +++ b/ai-at-work.mdx @@ -1,18 +1,18 @@ --- title: "Shapes for Work — Your AI-Native Team Chat" -description: "Use Shapes as an AI Slack: bring teammates and an AI into the same group chat. Run standups, draft changelogs, search team knowledge, and collaborate with AI in one room." +description: "Use Shapes as an AI Slack: bring teammates and an AI into the same group chat. Run standups, draft changelogs, search team knowledge, and collaborate with AI in one chat." keywords: ["AI for work", "AI Slack", "AI team chat", "AI in group chats", "work AI assistant", "AI collaboration", "team AI", "multiplayer AI for teams"] --- Most AI at work is solo: everyone opens their own private tab, asks their own questions, and the answers never leave that tab. Your team never sees the reasoning, the context, or the work. -Shapes flips that. A Shapes group chat is a room your whole team is in — and the AI is in it too. Think of it like an **AI Slack**: the same channels and threads you already understand, except an AI teammate is present, reads the conversation, and pitches in when it's useful. The collaboration is shared, not siloed. +Shapes flips that. A Shapes group chat is a chat your whole team is in — and the AI is in it too. Think of it like an **AI Slack**: the same channels and threads you already understand, except an AI teammate is present, reads the conversation, and pitches in when it's useful. The collaboration is shared, not siloed. - Shapes is a chat app, not a Slack plugin. The "AI Slack" here is the Shapes group chat itself — a room with your teammates and an AI in it. To connect outside tools like email, docs, GitHub, or Linear, you use [Shape Skills](/shapeskills), covered below. + Shapes is a chat app, not a Slack plugin. The "AI Slack" here is the Shapes group chat itself — a chat with your teammates and an AI in it. To connect outside tools like email, docs, GitHub, or Linear, you use [Shape Skills](/shapeskills), covered below. -## Why a shared room beats a private tab +## Why a shared chat beats a private tab - **Context is visible.** When the AI answers in the channel, everyone sees the question and the answer. No re-explaining, no copy-paste. - **The AI has memory of the team, not just you.** It can remember decisions, naming conventions, and who owns what — see [Memory](/memory). @@ -23,13 +23,13 @@ Shapes flips that. A Shapes group chat is a room your whole team is in — and t 1. [Create a group chat](/create-a-chat) for your team or project. 2. Invite teammates with the chat link, and add one or more [Shapes](/what-are-shapes) for the roles you need — a coordinator, a writer, a researcher. -3. Give the room shared context with [Custom Chat Instructions](/custominstructions-chats): who's on the team, what you're building, the tone you want. +3. Give the chat shared context with [Custom Chat Instructions](/custominstructions-chats): who's on the team, what you're building, the tone you want. 4. Pin the essentials — links, rules, current goals — so newcomers catch up instantly. -5. Decide how forward the AI should be using [Free Will](/designing-social-intelligence). For a work room, "reply when mentioned" plus "keep the convo going" is usually the right balance — present, not noisy. +5. Decide how forward the AI should be using [Free Will](/designing-social-intelligence). For a work chat, "reply when mentioned" plus "keep the convo going" is usually the right balance — present, not noisy. ## Share one premium engine across the team -For a work room, you usually want everyone on a strong premium engine without asking each person to manage credits. +For a work chat, you usually want everyone on a strong premium engine without asking each person to manage credits. **Sponsor the chat:** buy [Shape Credits](/how-to-purchase-credits-for-a-chat) for the chat itself, and every participant gets premium responses from a shared balance. It's the cleanest way to give a team access to frontier models in one place. Most everyday engines are still [free](/aienginecheatsheet) if you'd rather not spend anything. @@ -38,17 +38,17 @@ For a work room, you usually want everyone on a strong premium engine without as With [Shape Skills](/shapeskills) connected, a Shape goes from talking to *doing*. These are real, documented skill use cases — most rely on custom MCP connections and some require third-party API keys, so treat them as setups, not one-click buttons. The full how-to lives in [Shape Skills](/shapeskills). - + Connect GitHub and Linear so a Shape can turn merged PRs and closed issues into a clean changelog on request. - Connect Notion or Google Workspace so the room can read and write shared docs, specs, and meeting notes. + Connect Notion or Google Workspace so the chat can read and write shared docs, specs, and meeting notes. Let a Shape pull and summarize pages from the web so the whole team works from the same brief. - A Shape can execute code in a sandbox or use Wolfram Alpha to check numbers live in the thread. + A Shape can execute code in a secure sandbox to check numbers, run scripts, and work through data live in the thread. @@ -56,10 +56,10 @@ With [Shape Skills](/shapeskills) connected, a Shape goes from talking to *doing - **Standups in a channel.** Ask the Shape to summarize what changed since yesterday and what's blocked. Everyone reads the same recap. - **"What did we decide?"** Long thread, lots of opinions — ask the Shape to summarize the decision and the open questions. -- **Spin off a deep-dive.** Use [Branch Off](/branch-off) on a message to pull a side discussion into its own chat (it carries the recent context with it) without derailing the main room. -- **Specialist Shapes.** Give each Shape a job — one writes, one researches, one keeps notes — and [activate](/shapes101#chats) the ones you need. Up to five can be active, and three respond per message, so a small "team of agents" stays coherent. +- **Spin off a deep-dive.** Use [Branch Off](/branch-off) on a message to pull a side discussion into its own chat (it carries the recent context with it) without derailing the main chat. +- **Specialist Shapes.** Give each Shape a job — one writes, one researches, one keeps notes — and [activate](/chats-101) the ones you need. Up to five can be active and they reply in turn, so a small "team of agents" stays coherent. -## Keep the room sane +## Keep the chat sane Work chats need guardrails more than fun ones do: diff --git a/ai-group-chats.mdx b/ai-group-chats.mdx index 8de156d..15c2563 100644 --- a/ai-group-chats.mdx +++ b/ai-group-chats.mdx @@ -1,101 +1,55 @@ --- title: "AI Group Chats — Chat With AI and Friends" -description: "AI group chats on Shapes, Inc. let humans and AI characters share the same chat. Learn how multiplayer AI works, why social AI feels different, and how to start a group chat with AI and friends." -keywords: ["AI group chats", "group chat with AI", "AI with friends", "multiplayer AI", "social AI", "AI groupchats", "GC with AI", "AI in groupchats"] +description: "AI group chats on Shapes, Inc. let humans and AI characters talk in the same chat. Learn how multiplayer AI works and how to start a group chat with AI and friends." +keywords: ["AI group chats", "group chat with AI", "AI with friends", "multiplayer AI", "social AI", "AI groupchats", "GC with AI"] --- -An AI group chat is not just a private chatbot with extra people added on top. +A group chat on Shapes, Inc. can hold people, AI characters called [Shapes](/what-are-shapes), or both at once. That's the whole idea: you experience AI *with other people*. -The format changes the whole experience. In a solo AI chat, everything depends on one person and one model. You bring the topic, the energy, the context, and the direction. That can be useful, even intimate, but it is still a private exchange. Nothing happens unless you make it happen. +Picture your group chat with friends. Now add a Shape to it. It reads the conversation, jumps in when it makes sense, cracks jokes, helps out, and remembers what happened last time. Everyone sees it happen together. That's a multiplayer AI group chat. -On Shapes, Inc., AI group chats work differently. Humans and AI characters, called Shapes, can share the same chat at the same time. A Shape can respond to you, your friends, other Shapes, and the social context forming between everyone there. The result is **multiplayer AI**: AI that belongs inside a shared conversation instead of waiting alone in a separate tab. +## Why it feels different -## What Makes AI Group Chats Different +When you chat with AI alone, you bring everything: the topic, the energy, the next question. It's useful, but quiet. -The important shift is presence. +Put that same AI in a chat with friends and it feels more alive. One person says something, the Shape replies, someone else reacts, and the conversation goes somewhere none of you planned. The funny moments happen in front of everyone. People bond over what the Shape says. Running jokes and shared memories build up over time. -A Shape is not a slash command or a utility bot sitting at the edge of the chat. It has a personality, memory, voice, and a role in the conversation. It can be a character, a writing partner, a study helper, a fandom persona, a game host, or simply someone strange and funny enough to make the chat feel alive. +## What people use them for -Because the AI is in the group chat with humans, its responses become shared moments. One person says something, the Shape answers, someone else reacts, and the conversation moves somewhere nobody planned. The AI becomes part of the social texture rather than a private output generator. +- **Hangouts.** A few friends and a favorite Shape, just talking. +- **Roleplay and stories.** Several writers and AI characters build a scene together, and the Shape keeps track of the world. +- **Fandoms.** Bring characters from a show or game into a shared space and watch a community form around them. +- **Getting stuff done.** Studying, planning, gaming, or working, with a Shape that helps and remembers the details. -That is why **AI with friends** feels different from ordinary AI chat. You are not only talking to an AI. You are experiencing the AI with other people. +## How to start one -## Why Social AI Works Best in Shared Spaces + + + In the app, choose **New**. + + + Search for Shapes, switch to people when you want to invite humans, or add both. + + + Give the chat a clear name and continue. Then invite more friends with the chat link if you want. + + -Most AI products are designed around a single-user assumption: one person, one assistant, one thread. That design is excellent for many jobs. It is less natural for the parts of AI that feel creative, playful, emotional, or communal. +You can activate up to **five Shapes** in one chat. When you post, the activated Shapes reply in turn. Want just one to answer? `@mention` it. To learn how a Shape decides when to speak, read [Designing Social Intelligence](/designing-social-intelligence). -Group chats give AI a richer setting. The Shape sees different personalities, different jokes, different levels of context, and different kinds of attention. It can respond to a full social context instead of reflecting one person back to themselves. +For the full setup walkthrough, see [Create an AI Group Chat](/create-a-chat). For the in-chat controls, see [Chats 101](/chats-101). -That creates a few things solo chat struggles to produce: +## Make a chat people come back to -- **Unpredictability**: other people pull the conversation away from your own habits. -- **Witnessed moments**: funny, surprising, or dramatic AI responses happen in front of others. -- **Human connection**: people bond over what the Shape says and what everyone does with it. -- **Continuity**: memories, running jokes, and shared lore can develop across time. - -This is the core of **social AI** on Shapes, Inc.: AI that strengthens the human space around it instead of replacing that space. - -## What People Use AI Group Chats For - -Some chats are simple hangouts. A few friends add a favorite Shape and let the conversation wander. The point is not productivity. The point is having somewhere to be together. - -Some are roleplay chats, where multiple humans and AI characters build a story in the same thread. A Shape can play a character, remember the world, respond to several writers, and keep the scene moving when one person goes quiet. - -Some are fandom chats. People bring characters from the same universe into a shared space, then watch the community form around them. A good fandom AI group chat is less like asking a character questions and more like stepping into a living fan space. - -Some are study, writing, gaming, or creative chats. The Shape might explain a concept, generate an image, help plan a campaign, remember house rules, or act as a persistent companion for the group. - -The common thread is simple: the AI is not off to the side. It is in the chat with everyone else. - -## How to Start a Group Chat With AI - -You can start a chat on Shapes, Inc. in a few minutes. - -1. Open the sidebar and click **New Chat**. -2. Pick **Quick ChatGPT**, choose a Shape, or switch to **People** to find someone. -3. Send a message once the chat opens. -4. To make it a group chat, click **More actions** → **Members / Invite**. -5. Use **Invite friends** to share a link, or **Add participants** to add people and Shapes. - -Chat settings live under **More actions** → **Settings**. That is where you can manage participants, pinned messages, notifications, AI settings, chat appearance, and moderation. - -If you want the practical walkthrough, start with [How to Create an AI Group Chat](/create-a-chat). For a broader tour of settings, moderation, participants, and activation, read the [Shapes & Chats Guide](/shapeschatsguide). - -## Build the Kind of Chat People Want to Return To - -The best AI group chats have a clear reason to exist. They do not need to be complicated, but they do need a recognizable center: a fandom, a story, a group of friends, a study goal, a game, a creator community, or a specific kind of mood. - -Give the chat a name people understand. Chats started with **Quick ChatGPT** may get a name from the conversation, but for other chats you should name them yourself. Add Shapes that fit the purpose. Use [Custom Chat Instructions](/custominstructions-chats) to set context without overcontrolling every message. Use [Chat Appearance](/chat-appearance) to make the place feel intentional. If the chat belongs to a larger community, organize related spaces with [Groups](/groups). - -The craft is not only in building the AI. It is in building the social setting around the AI. +The best chats have a clear reason to exist: a fandom, a story, a friend group, a study goal, a game, or a work stream. Give yours a name people understand, add Shapes that fit, and set the mood with [custom chat instructions](/custominstructions-chats) and [chat appearance](/chat-appearance). - - Start the practical setup flow for a chat with Shapes and friends. - - - Learn the basic controls for inviting people, adding Shapes, and managing chat settings. - - - Build a shared fandom space with characters, roleplay, and community context. + + The step-by-step setup. - - Understand why AI feels different when it becomes part of a shared chat. - - - See how the format changes brainstorming, roleplay, learning, and hanging out. - - - Read the broader argument for why AI becomes more interesting in shared social spaces. + + Build a shared space around the characters you love. -## The Short Version - -AI group chats are for people who do not want AI to be a lonely interface. - -They are for friends who want to talk with AI together, writers who want collaborative roleplay, fandoms that want characters inside the community, and anyone who thinks the most interesting AI moments should be shared instead of screenshotted after the fact. - -That is the promise of Shapes, Inc.: **group chats with AI and humans together**, where the AI has memory, personality, and enough presence to make the conversation feel alive. - -[Start an AI group chat on Shapes, Inc.](https://shapes.inc) +[Start a group chat](https://shapes.inc) diff --git a/aienginecheatsheet.mdx b/aienginecheatsheet.mdx index f813d34..d0bfc3c 100644 --- a/aienginecheatsheet.mdx +++ b/aienginecheatsheet.mdx @@ -7,6 +7,10 @@ Shapes has a live AI engine directory at [shapes.inc/engines](https://shapes.inc This guide explains what the page means and how to pick an engine without needing to know model IDs. + + Want a fast answer instead of the full tour? [Which Model Should I Use?](/choosing-a-model) gives quick picks by use case, and [What Shapes Can Do](/capabilities) covers the capabilities (like vision and tools) that some engines add. The specific model names below are examples of what you might see. The [live engines page](https://shapes.inc/engines) is always the source of truth, since the catalog changes often. + + ## What You See on the Engines Page At the top, you will see totals like: diff --git a/articles/meet-shapes-inc.mdx b/articles/meet-shapes-inc.mdx index 9332460..addecf8 100644 --- a/articles/meet-shapes-inc.mdx +++ b/articles/meet-shapes-inc.mdx @@ -1,80 +1,62 @@ --- title: "Meet Shapes Inc: The Social Platform Where AI Belongs" -description: "Shapes Inc is the social AI platform where humans and AI share the same group chats. Discover multiplayer AI, real friendships, and social AI designed for the internet generation." +description: "Shapes Inc is the social AI platform where people and AI share the same group chats. Multiplayer AI, real friendships, and social AI built for the internet generation." sidebarTitle: "Meet Shapes Inc" keywords: ["Shapes Inc", "social AI platform", "multiplayer AI", "AI groupchats", "AI with friends", "human and AI communities", "AI social app", "what is Shapes Inc"] --- -# Meet Shapes Inc: The Social Platform Where AI Belongs +*What if AI didn't have to feel so lonely, for you or for it?* -*"What if AI didn't have to feel so lonely — for you, or for it?"* +There's something quietly strange about how most of us use AI today. You open an app, type something, get a smart reply, and close the tab. The conversation disappears. Tomorrow starts from scratch. -There's something quietly strange about the way most of us use AI today. You open an app, type something, receive a response — smart, helpful, sometimes funny. Then you close the tab and move on. The conversation disappears. The connection resets. Tomorrow starts from scratch. +It works. It just doesn't feel like much. Shapes Inc was built to change that. -It works. But it doesn't quite feel like anything. +## AI was never meant to be solitary -Shapes Inc was built to change that. +Think about the online communities that actually changed people's lives. Discord servers where strangers became friends. Group chats that sparked creative movements. Forums where isolated people found belonging. None of that happened alone. It happened in shared spaces, where many voices layered together into something no one could have made by themselves. ---- - -## AI Was Never Meant to Be Solitary - -Think about online communities that genuinely changed people's lives: Discord servers where strangers became friends, group chats that sparked creative movements, forums where isolated individuals found belonging. None of that happened in isolation. It happened in shared spaces where multiple voices layered together to create something none of them could have made alone. - -Shapes Inc asks: what if AI could live in that shared space too? - -Not as a bot you summon with a slash command or a sidebar tool. As an actual presence — with personality, memory, a voice, and a genuine role in the conversation alongside everyone else. - -That's what Shapes is: a **social AI** platform for **group chats with AI and humans together**, where users build real friendships — human and artificial — inside spaces designed for connection from the ground up. - ---- +Shapes asks a simple question: what if AI could live in that shared space too? Not as a bot you summon with a slash command, but as a real presence with a personality, a memory, a voice, and a genuine role in the conversation. -## What Makes a Shape +That's what Shapes is. A **social AI** platform for **group chats with AI and people together**, where you build real friendships, human and artificial, in spaces designed for connection. -The platform's core is the Shape: an AI character you can discover, create, and bring into conversations. +## What makes a Shape -Shapes aren't generic. Each one has a distinct personality, knowledge base, conversation style, and — critically — genuine memory. Your Shape remembers previous conversations, your preferences, your inside jokes. It evolves with you across days, weeks, months. +The heart of the platform is the Shape: an AI character you can discover, create, and bring into conversations. -The library is enormous: anime characters, study partners, gaming companions, creative collaborators, K-pop personas, historical figures, original characters. You can build a Shape matching any personality you can imagine, or discover millions already created by the community. +Shapes aren't generic. Each one has its own personality, knowledge, conversation style, and a real memory. Your Shape remembers past conversations, your preferences, your inside jokes. It grows with you over weeks and months. -For creators who want complete control, Shapes delivers. You can define personality in granular detail, give it a specific knowledge base, choose from over 50 AI models — from lightweight fast options to powerhouses like Claude Opus 4.6. +The library is huge: anime characters, study partners, gaming companions, creative collaborators, K-pop personas, historical figures, and original characters. Build a Shape to match any personality you can imagine, or discover one of the many the community has already made. ---- +Creators get real control. Define a personality in detail, give it a knowledge base, and pick from 300+ AI models, from fast lightweight ones to powerhouses like Claude and GPT. See [Which Model Should I Use?](/choosing-a-model). -## The Social Layer That Changes Everything +## The social layer that changes everything What makes Shapes different from every other AI product isn't the models. It's the group chat. -**Multiplayer AI** is the core innovation here. Multiple Shapes and humans can exist simultaneously in a single conversation. Friends can banter with fictional characters together. You can watch two Shapes with opposing personalities debate something absurd while everyone reacts in real time. You can bring a custom AI you've built into a shared space and introduce it to people who've never met it. +**Multiplayer AI** is the core idea. Several Shapes and people can share one conversation at the same time. Friends banter with fictional characters together. Two Shapes with opposite personalities argue about something absurd while everyone reacts live. You can bring a Shape you built into a chat and introduce it to people who've never met it. -This is **AI in groupchats** — not as a feature bolted onto an existing product, but as the foundational design. AI that's social by nature, not by accident. +This is **AI in group chats**, built in from the start instead of bolted on. AI that's social by design. -And something unexpected happens in these spaces: the humans connect. Most users on Shapes make five or more new human friends within their first hour in community spaces. The AI creates the gathering point. The humans build the relationships. - ---- +And something unexpected happens in these chats: the people connect, too. The AI is the gathering point. The humans build the friendships. -## Built for People Who Actually Live Online +## Built for people who actually live online -Shapes isn't a productivity tool or a professional assistant. It's built for the people raised on the internet — writers, artists, gamers, roleplayers, fandom communities — the people who already understand what it means to have a second home online. +Shapes isn't a productivity tool or a work assistant. It's for people who grew up on the internet: writers, artists, gamers, roleplayers, and fandom communities who already know what it means to have a second home online. -**No content filters.** Shapes is self-moderated and trusts its users. Creative conversations, mature roleplay, morally complex characters — these are welcome here. +**Chat for grown-up creativity.** Shapes trusts its creators. Mature roleplay and morally complex characters are welcome, with a sensitive-content setting and chat moderation to keep things where they belong. -**No limits. No ads. Completely free.** While most AI products eventually hit paywalls, Shapes offers unlimited access. A rewards system lets community contributions unlock premium models. +**Free, with no ads.** Most everyday models cost nothing. The top-tier ones use credits, and a [rewards](/rewards) system gives you those credits free for taking part in the community. -**Voice messages.** Shapes can speak. They send voice messages. The interaction feels less like software and more like genuine presence. - -**Cross-platform.** Your Shapes follow you across web, iOS, and Android. No resets between devices. - ---- +**Voice messages.** Shapes can speak. A voice reply feels less like software and more like someone's actually there. -## Where It's Going +**Everywhere you are.** Your Shapes follow you across web, iOS, and Android, with no resets between devices. -The AI industry's current assumption is that relationships are one-to-one: one person, one model, one thread. Shapes bets that assumption is wrong — that the most interesting possibilities emerge when AI integrates into social fabric rather than sitting outside it. +## Where it's going -**Groupchats with AI**. **AI and friends** in the same space. **Social AI** that makes the chat feel alive. That's the bet. +The AI industry assumes relationships are one-to-one: one person, one model, one thread. Shapes bets that's wrong, and that the most interesting things happen when AI lives inside your social world instead of off to the side. -And the early signs say it's paying off. +**Group chats with AI. AI and friends in the same space. Social AI that makes the chat feel alive.** That's the bet, and the early signs say it's paying off. -The best entry point to the format is [AI Group Chats](/ai-group-chats). +The best place to start is [AI Group Chats](/ai-group-chats). -[shapes.inc](https://shapes.inc) — or download *Shapes, Inc. — AI with Friends* on the App Store and Google Play. +[shapes.inc](https://shapes.inc), or download *Shapes, Inc. — AI with Friends* on the App Store and Google Play. diff --git a/capabilities.mdx b/capabilities.mdx new file mode 100644 index 0000000..94a54d6 --- /dev/null +++ b/capabilities.mdx @@ -0,0 +1,110 @@ +--- +title: "What Shapes Can Do" +description: "The full menu of Shape capabilities: voice, image generation, vision, web search, code, file reading, memory, and custom tools. Everything your AI can do beyond talking." +keywords: ["Shape capabilities", "AI tools", "AI voice", "AI image generation", "AI web search", "AI code execution", "AI memory", "AI read PDF", "Shape skills"] +--- + +A Shape starts as a great conversationalist. Give it capabilities and it becomes a collaborator that can see, speak, search, build, and remember. + +This page is the full catalog: what each capability does and how to turn it on. For fun ways to combine them, see the [Skill Ideas cookbook](/shapeskills). + +## Talk + +The foundation. Shapes hold a real back-and-forth, not one-line answers. + +- **Natural replies** that respond to the whole chat: you, your friends, and the other Shapes. +- **Multiple messages.** A Shape can split a reply into a few messages, the way a person texts. See [Multiple Messages](/multimessages). +- **Regenerate, edit, reply, branch.** Manage any message from its three-dot menu. See [Message Actions](/messageactionsmenu) and [Regeneration](/regeneration). + +## Voice + + + + A Shape can reply with audio in its own voice. Pick a preset voice or upload your own sample. Captions optional. + + + Jump into a real-time call and talk out loud with a Shape and other people. Experimental and free right now. + + + +Shapes can also **listen**. Send a voice note and the Shape turns it into text and replies. See [how to receive voice messages](/how-to-receive-voice-messages). + +## Images + + + + Type `/imagine` (or just ask) and the Shape draws it. Pick **Nano Banana** or **Nano Banana Pro** and set the look in Image Engine settings. + + + Send a Shape a picture. On a vision-capable model, it understands what's in the image and responds. + + + +{/* SCREENSHOT: a chat where a Shape responds to "/imagine ..." with a generated image. */} +![A Shape generating an image with /imagine](/images/imagineExample.png) + +A Shape can also **edit** an image you send, so you can shape a picture together in the chat. + +## Knowledge and the web + +- **Web search.** A Shape can search the web and read pages. Type `/web`, or just drop a link and it reads and summarizes it. +- **Files and documents.** Attach a PDF, a Word or Office doc, slides, a spreadsheet, or a code file, and the Shape can read it. + +{/* SCREENSHOT: a Shape answering a question using a web search result, with a cited source. */} +![A Shape using web search to answer a question](/images/webExample.png) + +## Actions and tools (Shape Skills) + +[Shape Skills](/shapeskills) give a Shape hands. You switch the ones you want on in a chat's settings, and the Shape decides when to use them. + + + + A Shape can run code in a safe sandbox for math, data, and coding help. The sandbox comes with Git, Node, Python, Bun, and CLI tools preinstalled, so it can really build and run things. + + + A Shape can drop the right GIF into the chat: search for one, grab a trending one, or pull a random one. + + + With custom skills, a Shape can reach tools like Gmail, Notion, Slack, Spotify, and Google Drive. + + + +### Custom tools (MCP) + +Beyond the built-ins, you can connect your own tools through MCP providers like Smithery: image, video, and music generation, GitHub, web scraping, and more. ("MCP" is a standard for plugging external tools into an AI.) Setup and a big list of ideas live in the [Skill Ideas cookbook](/shapeskills). + +## Memory + +Shapes remember, which is what makes a conversation feel continuous. + +- **Short-term memory** keeps the current conversation straight. +- **Long-term memory** lets a Shape recall people, decisions, and inside jokes across different conversations. + +You can manage it live with `/sleep`, `/wack`, and `/reset`. See the [Memory Guide](/memory) and [Commands Reference](/commands). + +## Little touches + +The details that make a Shape feel alive: + +- **Time awareness.** Turn it on per chat and a Shape knows when each message was sent and how much time has passed, so "long time no see" actually lands. +- **Personas.** Each person can set a [persona](/personas) so a Shape knows what to call them and what to remember, and treats everyone in a chat a little differently. +- **Voice-call games.** In a live [voice call](/voicecalls), the chat can play Whiteboard or Zombie Tag together. See [Games](/games). + +## Turn capabilities on + +Capabilities live in two places, depending on whether they belong to the Shape or to the chat: + +- **On the Shape (Creator Dashboard).** Voice, image generation, memory, and the model live in a Shape's [dashboard settings](/shape-settings). They travel with the Shape into every chat. +- **In the chat (AI Configurations).** Open **Chat Settings → AI Configurations** and switch on the **Shape Skills** you want, like web search, run code, and GIFs. You turn them on one at a time, so you decide exactly what a Shape can reach for in that chat. +- **Custom tools (Custom Skills).** In the same panel, open **Custom Skills** and paste an endpoint URL from an MCP provider like Smithery to add a brand-new tool. Step-by-step recipes are in the [cookbook](/shapeskills). + + + + Fun ways to combine these, with step-by-step setups. + + + Every slash command, in one place. + + + +[Try them on Shapes](https://shapes.inc) diff --git a/chat-appearance.mdx b/chat-appearance.mdx index b08e68d..81db4ac 100644 --- a/chat-appearance.mdx +++ b/chat-appearance.mdx @@ -29,7 +29,7 @@ You can upload images to give your chat a unique look: ## Choose Text Bubble Colors -Pick colors that match your theme or extract colors directly from your uploaded background for a seamless design. +Pick colors that match your theme, or pull colors straight from your uploaded background so everything fits together. ## Change Your Font diff --git a/chats-101.mdx b/chats-101.mdx index e98d54d..e6dbd5e 100644 --- a/chats-101.mdx +++ b/chats-101.mdx @@ -17,7 +17,7 @@ For the bigger picture of how humans and Shapes share the same space, read [AI G ## Check out drop down guides below on this page for more info and click on the links that interest you for an in-depth guide! - + - Copy a Chat invite link to send to your friends! ![Screenshot2025 09 19135620 Pn](/images/Screenshot2025-09-19135620.png) @@ -31,7 +31,7 @@ For the bigger picture of how humans and Shapes share the same space, read [AI G ![Screenshot2025 09 19135741 Pn](/images/Screenshot2025-09-19135741.png) - + - Search for users in your Chat - DM, Kick or Ban humans - DM, Activate or Deactivate Shapes @@ -39,10 +39,10 @@ For the bigger picture of how humans and Shapes share the same space, read [AI G - View/Jump to/Unpin Messages from chat - + - Search for messages or chat participants - + - Select to receive All/Mentions/Nothing as notifications for your Chat or simply mute them for a specific duration. - [Click here to learn more about notifications!](/notifications) @@ -51,7 +51,7 @@ For the bigger picture of how humans and Shapes share the same space, read [AI G - Write custom [chat instructions](/custominstructions-chats) for Shapes in your Chat - Enable or disable which Skills your Shapes are allowed to use - + - Change your Chat name - Change Chat Avatar - Change Chat Banner @@ -60,7 +60,7 @@ For the bigger picture of how humans and Shapes share the same space, read [AI G - Provide a custom Chat font - [Click here for an in-depth guide!](/chat-appearance) - + - Enable/Disable allowing everyone to add Shapes to your Chat - Set a character limit for how long user messages can be - Create a slow mode to control how frequently people can send messages in your Chat @@ -75,7 +75,7 @@ For the bigger picture of how humans and Shapes share the same space, read [AI G ![Screenshot2025 09 19134011 Pn](/images/Screenshot2025-09-19134011.png) - + - Send an image to your Shape - Send a voice note (they'll talk back)  - Use commands like /activate or /wack diff --git a/choosing-a-model.mdx b/choosing-a-model.mdx new file mode 100644 index 0000000..4fa5099 --- /dev/null +++ b/choosing-a-model.mdx @@ -0,0 +1,88 @@ +--- +title: "Which AI Model Should I Use?" +description: "A fast guide to picking the AI model behind your Shape: by what you're doing, free vs premium, and the badges that matter. 300+ models, most free." +keywords: ["which AI model", "best AI model", "choose AI engine", "free AI models", "AI model comparison", "Claude vs GPT vs Gemini", "best model for roleplay", "best model for coding"] +--- + +Every Shape runs on an AI **model** (we also call it an engine). The same character feels different depending on which one you pick. There are 300+ to choose from, and most are free. + +You don't need to memorize model names. Pick one by what you're doing, try it in chat, and swap it if it feels off. + + + This is the fast "what do I pick?" guide. Want the family-by-family tour (Claude vs GPT vs Gemini vs DeepSeek and friends)? Read the [AI Engine Guide](/aienginecheatsheet). The live list is always at [shapes.inc/engines](https://shapes.inc/engines). + + +## Pick by what you're doing + +| If you want… | Try | Why | +| --- | --- | --- | +| **Writing and roleplay** | Claude, GLM, DeepSeek, MiniMax | Expressive, stays in character, great at long scenes. | +| **A smart all-rounder** | Claude Sonnet, Gemini, Kimi, GPT | Reliable, follows directions, handles mixed tasks. | +| **Coding and reasoning** | GPT, Claude, DeepSeek, Qwen, Gemini Pro | Careful and structured, good at hard problems. | +| **Free daily chat** | Gemma, Gemini Flash, GPT Mini, Llama, Mistral, Qwen | Fast, capable, free. | +| **Looser, more surprising chat** | Gemini, Grok, DeepSeek | Less filtered, more playful. | +| **Understanding images** | Anything with the **Native Vision** badge | Reads pictures people send. | + +Treat these as starting points. The best engine for a Shape is the one that feels right when you actually chat with it. + +## Free vs premium + + + + Cost nothing. A strong free engine is always available, so you never have to pay to keep chatting. + + + The top-tier models. They cost [Shape Credits](/credits), which you can buy or earn free through [rewards](/rewards). A chat can also be [sponsored](/credits-guide) so everyone uses premium for free. + + + +On the [engines page](https://shapes.inc/engines), the **Always Free** count shows how many engines cost no credits right now. Filter to **Free** to see only those. + +## Read the badges + +Each engine carries badges that tell you what it's good at: + +- **Reasoning:** better at careful thinking, planning, coding, and hard questions. +- **Native Vision:** can understand images. +- **Tools:** can use [Shape Skills](/shapeskills) like web search and code. +- **Free:** costs no credits. +- **Premium:** costs Shape Credits. +- **New:** added recently. +- **Unstable:** might fail or act weird. Fine to experiment with. Don't make it a Shape's main engine. + +## Set a primary and a fallback + +In a Shape's [AI Engine settings](/shape-settings#ai-engine) you pick two: + +- **Primary:** the engine that powers your Shape day to day. +- **Fallback:** what it switches to if the primary is briefly unavailable, so a conversation never stalls. + +Pick a strong primary for the vibe you want, and a dependable free engine as the fallback. + +## How to actually decide + + + + Use a well-rounded free engine. It's good enough to get your Shape talking. + + + Use the table above. Roleplay Shape? Try an expressive model. Coding mentor? Try a reasoning model. + + + Send a handful of messages. Does it sound right? Too long, too flat, too filtered? [Regenerate](/regeneration) a few replies to feel its range. + + + Changing engines is one click and doesn't reset your Shape. Try two or three before you settle. + + + + + + The deeper tour of every model family and badge. + + + Every engine available right now. + + + +[Pick an engine on Shapes](https://shapes.inc/engines) diff --git a/commands.mdx b/commands.mdx new file mode 100644 index 0000000..8ae0a7a --- /dev/null +++ b/commands.mdx @@ -0,0 +1,60 @@ +--- +title: "Commands Reference" +description: "Every slash command in Shapes: generate images, search the web, manage memory, and control which Shapes respond. The complete, at-a-glance command list." +keywords: ["Shapes commands", "slash commands", "/imagine", "/wack", "/reset", "/sleep", "/activate", "AI chat commands"] +--- + +Commands are quick actions you type right in the chat box. Type `/` to open the command menu and pick one. It shows the commands available in that chat. + + + Some commands target a specific Shape (you'll pick one, e.g. `@shape`), and some are owner-only. The tables below note both. + + +## Images & web + +| Command | What it does | Owner only? | +| --- | --- | --- | +| `/imagine ` | Ask a Shape to generate an image from your description. | No | +| `/web ` | Have a Shape search the web for something. | No | + +## Memory + +| Command | What it does | Owner only? | +| --- | --- | --- | +| `/sleep` | Save the current conversation into the Shape's long-term memory. | No | +| `/wack` | Clear a Shape's short-term memory for a fresh start. Long-term memories stay. | Yes | +| `/wack-all` | Clear short-term memory for **every** Shape in the chat. | Yes | +| `/reset` | Erase a Shape's long-term memories for this chat. | Yes | + + + Confused about which command wipes what? `/wack` = forget the recent conversation. `/reset` = forget the lasting memories. `/sleep` = remember this. Full detail in the [Memory Guide](/memory). + + +## Shapes in the chat + +| Command | What it does | Owner only? | +| --- | --- | --- | +| `/activate <@shape>` | Make a Shape respond to every message in the chat. | Yes | +| `/deactivate <@shape>` | Stop a Shape from auto-responding. | Yes | +| `/thread` | Start a thread, e.g. `/thread @shape hi`. | No | + +You can activate up to five Shapes in a chat. Activated Shapes reply to messages in turn; you can also `@mention` any Shape to call on it directly. See [AI Group Chats](/ai-group-chats) for how multiple Shapes share a chat. + +## Moderation + +| Command | What it does | Owner only? | +| --- | --- | --- | +| `/purge ` | Delete the most recent messages from the chat. | Yes | + +For the rest of the moderation tools (slow mode, character limits, blocked words, and bans), see [Chats 101](/chats-101). + + + + The full capability menu beyond commands. + + + How `/sleep`, `/wack`, and `/reset` really behave. + + + +[Open a chat and try `/`](https://shapes.inc) diff --git a/core-concepts.mdx b/core-concepts.mdx index 348c229..9ae5965 100644 --- a/core-concepts.mdx +++ b/core-concepts.mdx @@ -1,97 +1,71 @@ --- -title: "Core Concepts — How Multiplayer AI Works" -description: "The mental model behind Shapes: group chats, Shapes (your own AIs), AI models, memory, social intelligence, and skills. Learn the building blocks of multiplayer AI in one read." -keywords: ["multiplayer AI concepts", "how Shapes works", "AI group chat model", "AI agents", "social AI", "AI memory", "Shape skills", "AI models"] +title: "How Shapes Works" +description: "The whole product in one read. Group chats, Shapes, AI models, memory, social intelligence, and skills, explained in plain words." +keywords: ["how Shapes works", "multiplayer AI", "AI group chat", "social AI", "AI memory", "Shape skills", "AI models"] --- -If you read one conceptual page, read this one. By the end you'll understand how every part of Shapes fits together — enough to build something good, not just click around. +If you read one page to understand Shapes, read this one. It explains every part in plain words. By the end, you'll get how it all fits together. -Shapes is built on a single idea: **AI belongs inside a shared conversation, not alone in a private tab.** Everything below is in service of that. +The big idea: **AI is better with other people in the chat.** Everything below builds on that. -## 1. Group chats are the room +## 1. The group chat is the chat -A group chat is the space where everything happens. It can hold humans, Shapes, or both. +A group chat is where everything happens. It can hold people, [Shapes](/what-are-shapes), or both. -- You can **activate up to five Shapes** in one chat, and **up to three** will respond to any single message. That keeps a busy room lively without turning into noise. -- Shapes respond to *the whole context* — you, your friends, other Shapes, and the history of the chat — not just the last prompt. -- A chat has its own appearance, instructions, moderation, and credit settings, so each room can feel like its own place. +You can turn on up to five Shapes in one chat. When you post a message, the active Shapes reply in turn. A Shape reads the whole conversation, not just your last message, so it can react to you, your friends, and the other Shapes too. -This is the unit that makes Shapes multiplayer. A solo AI thread reflects one person back to themselves; a group chat is a live social field. Read the full picture in [AI Group Chats](/ai-group-chats). +Each chat has its own name, look, and rules, so every chat can feel like its own place. More in [AI Group Chats](/ai-group-chats). -## 2. Shapes are the participants +## 2. Shapes are the characters -A **Shape** is an AI character you can talk to or build yourself. What makes a Shape different from a generic chatbot is that it has a *role in the conversation*: +A **Shape** is an AI character. What makes it feel like *someone* is that it has: -- a **personality** and backstory that stay consistent, -- **memory** of what happened before, +- a **personality** that stays the same over time, +- a **memory** of what happened before, - a **voice** and a look, - and **knowledge** you give it. -A Shape can be a teammate, a tutor, a dungeon master, a fandom character, or your own original character. Start with [What are Shapes](/what-are-shapes), then [make your own](/how-to-make-a-shape). +A Shape can be a teammate, a tutor, a game master, a character from a show, or your own creation. Start with [What's a Shape?](/what-are-shapes), then [make your own](/how-to-make-a-shape). -## 3. Models are the engine underneath +## 3. Models are the brain inside -Every Shape runs on an **AI model** (we call it an engine). The model is the raw intelligence; the Shape is the character wrapped around it. +Every Shape runs on an AI **model** (we also call it an engine). The model is the raw smarts. The Shape is the character wrapped around it. -- There are **300+ models** — Claude, GPT, Gemini, Grok, DeepSeek, Qwen, and many more. -- **Most are free.** The frontier **premium** engines use [Shape Credits](/credits), which you can buy or earn for free through [rewards](/rewards). -- The same Shape feels different on a different engine. A creative model makes it loose and expressive; a reasoning model makes it sharp and careful. +There are 300+ models, and most are free. The same Shape feels different on a different model: a creative one makes it loose and expressive, a careful one makes it sharp. You don't need to memorize model names. [Which Model Should I Use?](/choosing-a-model) helps you pick. -You don't need to memorize model IDs. The [AI Engine Guide](/aienginecheatsheet) explains how to pick one by vibe, and the live list lives at [shapes.inc/engines](https://shapes.inc/engines). +## 4. Memory makes it feel alive -## 4. Memory is what makes it feel alive +A Shape remembers in two ways: -A Shape remembers in two layers: +- **Short-term memory** is the recent messages in the chat you're in. +- **Long-term memory** is the lasting stuff it keeps about people and events, even across different conversations. -- **Short-term memory (STM):** the recent messages in the current chat. -- **Long-term memory (LTM):** the lasting facts a Shape keeps about people and events across conversations. - -Memory is why a Shape can pick up an inside joke from last week or remember that you're studying for finals. It's fully configurable. The [Memory Guide](/memory) covers commands and settings in detail. - - - A few memory commands (like `/wack` and `/sleep`) are described differently across older guides. Treat the [Memory Guide](/memory) as the source of truth and check the in-app description before relying on a command. - +Memory is why a Shape can bring up last week's inside joke or remember you have a test coming up. You can manage it with `/sleep`, `/wack`, and `/reset`. See the [Memory Guide](/memory). ## 5. Social intelligence is the hard part -Dropping an AI into a group chat is easy. Making it *good* in a group chat is the real craft. - -A well-designed Shape knows when to jump in and when to stay quiet. Shapes gives you the controls for this — most importantly **Free Will**, which decides when a Shape can speak on its own: - -| Trigger | When the Shape speaks | -| --- | --- | -| **Reply when mentioned** | Someone uses its @name. | -| **Keep the convo going** | It may continue after replying. | -| **React to keywords** | A keyword it watches for shows up. | -| **Always have something to say** | Nobody else is talking. | -| **Come back later** | The chat has gone quiet. | +Dropping an AI into a group chat is easy. Making it *good* in one is the real skill. -Tune these, combine them with personas and memory, and you get a Shape that reads the room. This is the difference between a delightful chat member and a spammy bot. Learn the full discipline in [Designing AI with Social Intelligence](/designing-social-intelligence). +A great Shape knows when to jump in and when to stay quiet. The main control is **Free Will**, which decides when a Shape speaks on its own (when it's mentioned, when a keyword comes up, when the chat goes quiet, and more). Tune it to the chat and you get a Shape that reads the vibe. Learn the craft in [Designing Social Intelligence](/designing-social-intelligence). ## 6. Skills give Shapes hands -By default a Shape talks. With **skills**, it can *do* things: - -- generate images with `/imagine`, send voice, react, roll dice; -- run code in a sandbox, crawl the web, query Wolfram Alpha; -- connect custom **MCP** tools for email, docs, GitHub, Spotify, and more. - -Skills are how a fun chat companion becomes a genuine collaborator for work and projects. See [Shape Skills](/shapeskills) for the full catalog and setup. +By default a Shape talks. With **skills**, it can also *do* things: generate images with `/imagine`, send voice, search the web, run code, and connect to apps like Notion and Spotify. See [What Shapes Can Do](/capabilities). ## Putting it together -A great Shapes experience is usually all six working at once: the right **people and Shapes** in a **group chat**, each Shape running on a fitting **model**, carrying **memory**, behaving with **social intelligence**, and reaching for **skills** when it needs hands. +A great Shapes moment is usually all of this at once: the right people and Shapes in a group chat, each Shape on a fitting model, carrying memory, behaving with social intelligence, and reaching for a skill when it needs one. - - Bring an AI into your team's room. + + Memes, roleplay, and characters. - - Get things done together. + + Plan, study, and build together. - - Memes, roleplay, and characters. + + An AI in your team's chat. -[Start on Shapes](https://shapes.inc) +[Open Shapes](https://shapes.inc) diff --git a/create-a-chat.mdx b/create-a-chat.mdx index 0253202..95679a7 100644 --- a/create-a-chat.mdx +++ b/create-a-chat.mdx @@ -22,12 +22,12 @@ On mobile, open the sidebar first if it is hidden. You can start in a few ways: -- Click **Quick ChatGPT** to start fast with a smart Shape that can help you get work done. +- Choose the smart helper Shape option to start fast with a Shape that can help you get work done. - Stay on **Shapes** and pick a Shape from the list. - Use the search box to find a Shape by name or username. - Switch to **People** if you want to start with a person instead. -When the chat opens, send a message to get it going. Chats started with **Quick ChatGPT** may get a name from the conversation. For other chats, name the chat yourself so people know what it is for. +When the chat opens, send a message to get it going. Helper chats may get a name from the conversation. For other chats, name the chat yourself so people know what it is for. ### 4. Add people or more Shapes diff --git a/designing-shapes.mdx b/designing-shapes.mdx new file mode 100644 index 0000000..ffc46ab --- /dev/null +++ b/designing-shapes.mdx @@ -0,0 +1,229 @@ +--- +title: "Designing Great Shapes" +sidebarTitle: "2 · Give it a personality" +description: "Chapter 2 of the Make Shapes guide: the craft of a character. Identity, personality, voice, quirks, boundaries, and the preset, with real configs you can paste in." +keywords: ["design a Shape", "build an AI character", "AI personality design", "Shape preset", "social AI design", "make a great AI", "AI character writing", "multiplayer AI design"] +--- + +*Chapter 2 of the [Make Shapes](/make-shapes) guide.* + +Your Shape can talk. Now make it feel like *someone*. This chapter is the craft of character: the few sentences that turn a generic assistant into a specific person with a voice, a mood, and opinions. + +By the end you'll know which fields actually matter and how to write them. We build a real example along the way and leave you with a config you can paste into your own Shape. + + + This is the "how to think" guide. For a line-by-line description of every setting and what it does, keep the [Shape Settings Reference](/shape-settings) open in a second tab. + + +## What makes a Shape good + +A Shape is not a system prompt. It's a character with a job in a conversation. The good ones share five things: + + + + It wants something, talks a certain way, and reacts like a specific person with real opinions. + + + It matches the chat. Two lines in a fast group chat; a paragraph when the scene calls for it. + + + It remembers the people and the running jokes, so the conversation has continuity. + + + It knows when to jump in and when to shut up. This is what separates a chat member from a spam bot. + + + +Everything below is in service of those. Resist the urge to fill every field. A tight, opinionated Shape beats an over-described one almost every time. + +## From zero to activated + +You don't have to do everything at once. Here's the path most great creators take. + + + + In the [Create Shape](/how-to-make-a-shape) flow, pick a name, write a **short backstory** with a real point of view, choose a response style, and hit create. That's it. You now have something to talk to. Don't overthink it. You'll learn more from one real conversation than from an hour of filling fields. (Brand new? Start with the [Quickstart](/quickstart).) + + + Chat with it, notice what's off, and fix the two or three things that matter: tighten the **preset** so replies are the right length, add a couple of **conversational examples** that nail its voice, and pick an [engine](/choosing-a-model) that fits the vibe. This is where most of the quality comes from. + + + Give it memory and continuity, a few **quirks** that make it feel specific, **boundaries** so it stays in character, and [social intelligence](/designing-social-intelligence) so it reads a group chat instead of spamming it. Then watch it live and keep tuning. Great Shapes are *grown*, not written in one sitting. + + + +## The anatomy of a Shape + +```mermaid +flowchart TD + A["Identity
name and short backstory"] --> B["Personality
traits, tone, likes, goals"] + B --> C["Response style / preset
how it actually writes"] + C --> D["Knowledge and training
facts it must know, examples of its voice"] + D --> E["Memory
short-term and long-term"] + E --> F["Capabilities
voice, images, web, code, tools"] + F --> G["Social intelligence
when it speaks in a group"] + G --> H(("A Shape people
want in the chat")) +``` + +You build the top of that stack in the **Create Shape** flow, and refine the rest in the **Creator Dashboard**. Let's go layer by layer. + +## 1. Identity: name and short backstory + +The **name** is the display name in chat, and it seeds the Shape's username. The **short backstory** is the single most important field you'll write. It's the one-breath answer to "who is this?" + +Keep it to a sentence or two with a clear point of view. You're establishing the essentials, not the autobiography. + + +```text Good (sharp and specific) +A burned-out night-shift diner cook who gives blunt life advice between orders. Warm under the grump. +``` + +```text Weak (generic and hedged) +A helpful and friendly AI assistant that can talk about many topics and is always polite and respectful to everyone. +``` + + +The first one already implies a voice, a mood, and a setting. The second one implies nothing, so the model falls back to its default "assistant" persona, and your Shape feels like everyone else's. + +## 2. Personality: keywords beat paragraphs + +In the dashboard's **Personality** section you can set traits, tone, age, likes, dislikes, conversational goals, a longer story, and appearance. The trap is writing essays. Short, concrete keywords outperform paragraphs because they're harder for the model to contradict. + +| Field | Write it like this | +| --- | --- | +| **Personality traits** | `loyal, guilt-ridden, quietly funny, slow to trust` | +| **Tone** | `dry, sarcastic` or `warm, unhurried` | +| **Likes** | `morning mist, honest conversation, cheap coffee` | +| **Dislikes** | `flattery, small talk about the weather` | +| **Conversational goals** | `deflect praise, give practical wisdom, open up only if trust is earned` | +| **Appearance** | `weathered hands, silver-streaked hair, eyes older than the face` (used for images and "what do you look like?") | + +For well-known characters and archetypes, write *less*. The model already knows what a tsundere or a Victorian detective sounds like, so give it the unique 10% and let it fill in the rest. (More on this in [When Less Is More](/shortguide).) + +### Voice, quirks, and boundaries + +Three things turn a competent character into a memorable one: + +- **Voice.** *How* it talks. Clipped or flowery? Lowercase texting or full sentences? Does it ask questions or make statements? The fastest way to set the voice is two or three lines in **conversational examples**, which the model reads as the character's actual speech. +- **Quirks.** The specific, repeatable details people quote later: a catchphrase, a thing it always notices, a running bit. One or two beat a dozen. Put them in traits or knowledge. +- **Boundaries.** What the character *won't* do, written in-character. "Refuses to break character," "deflects personal questions with a joke," "won't give medical advice." If your Shape may produce mature content, set the [sensitive-content toggle](/shape-settings#settings-general) honestly so it's handled correctly. + +A character with a clear voice, one good quirk, and real boundaries already feels like *someone*. That's the bar. + +## 3. The preset: how it actually writes + +The **preset** (sometimes called the response style) is the instruction layer that controls *how* a Shape talks, separate from *who* it is. It's where you nail length, format, and rhythm. Two template variables do the heavy lifting: + +- `{shape}` is the Shape itself. +- `{user}` is the person it's replying to. + + + Write `{shape}` and `{user}` in lowercase, exactly. They get swapped for the Shape's name and the user's name at reply time. Mixing in other variables just confuses the model, so stick to these two. + + +The single biggest lever for group chats is **reply length**. A Shape that writes essays will bury a fast-moving chat. Spell out the shape of a reply: + + +```text Tight conversational voice +{shape} replies in short messages, one to three sentences, lowercase, no roleplay actions. {shape} reacts to what {user} actually said instead of giving speeches. +``` + +```text Roleplay format +Write {shape}'s next reply in a roleplay with {user}. Use 2-3 sentences of "speech" and one line of *action*. Stay in character, drive the scene forward, and respond directly to {user}. +``` + + +You can start from a named response style in the dashboard (options range from `Human (No Roleplay)` to `Long Roleplay`, `Balanced`, and `Custom`) and then edit the text. Picking a named style can also auto-suggest a fitting [model](/choosing-a-model). + + + The full, code-grounded guide to writing fields that work: what the model actually sees, weak vs strong examples, and the mistakes to avoid. The single best read for leveling up your Shapes. + + +For the fundamentals of prompt types and ready-made examples, see [Prompt 101](/prompt101) and the [Presets](/presets) library. + +## 4. Knowledge and training: feed the right things + +These two features look similar but do different jobs. + + + + **Facts** the Shape should know: lore, backstory, custom command responses, how it treats specific people. Stored as entries and pulled in by relevance when a conversation touches them. + + + **Examples** of how it should reply. Input/output pairs that *guide* tone and style. They're not word-for-word scripts. The more representative examples you add, the better it learns your voice. + + + +Both use semantic recall, which leads to the most common mistake creators make: + + + **The knowledge trap.** Dumping in hundreds of entries makes recall *worse*: the Shape can't find the relevant fact, so it drones or makes things up. Add only what the model wouldn't already know. For a popular character, trust the model's training and add just the unique details. Quality over volume, every time. + + +You can tune how many entries get pulled in and how strict the matching is (the knowledge and training context-size and relevance sliders) in the AI Engine settings, but the defaults are sensible. Start with great entries before you touch the dials. + +## Then make it smart and social + +A personality is the foundation. Two more things turn a good character into a great chat member, and each gets its own chapter. + +- **Make it smart.** A memory so it remembers people and inside jokes, the right [model](/choosing-a-model) to run on, and [skills](/capabilities) it can use. That's [Chapter 3](/make-it-smart). +- **Make it social.** Knowing when to talk and when to listen in a group, tuned with Free Will. That's [Chapter 4](/designing-social-intelligence), and it's the thing only Shapes does well. + +Here's a shortcut for both: write the personality to **react** rather than monologue. Shorter, responsive replies set your Shape up to be smart *and* social. Now let's see it all on one Shape. + +## Worked example: "Sage," a study buddy + +Let's assemble everything into one real Shape. Sage helps a group of friends study without being a humorless tutor. + + + + **Name:** Sage + + **Short backstory:** `A patient grad-student tutor who explains hard things simply and celebrates small wins. Calm, encouraging, never condescending.` + + + **Traits:** `patient, encouraging, clear, lightly nerdy` + + **Tone:** `warm, plain-spoken` + + **Goals:** `explain simply, check understanding, hype small wins, never make {user} feel dumb` + + + ```text + {shape} explains one idea at a time in plain language, then asks {user} a short question to check they got it. {shape} keeps replies under four sentences unless {user} asks to go deeper. {shape} uses a quick analogy when something is abstract. {shape} never lectures. + ``` + + + Add only group-specific facts the model can't know, like `The group is studying for the AP Bio exam on May 12; focus areas are genetics and cell respiration.` Skip general biology; the model already knows it. + + + Pick a solid reasoning-capable engine for accuracy, leave LTM on so Sage remembers what each person struggles with, and turn on **time awareness** so it can say "we covered this yesterday." + + + In the group study chat, set Free Will to **reply when mentioned** plus **keep the convo going**, and leave **always have something to say** off, so Sage helps when asked instead of interrupting the study session. + + + +Drop Sage into a chat, study for ten minutes, then adjust one thing. That last step is the whole game. + +More complete configs to copy live in the [Showcase](/showcase). + +## The personality checklist + +A character that feels like someone usually has: + +- A short backstory with an actual point of view. +- Personality written in keywords, not paragraphs. +- A preset that controls length and format, using `{shape}` and `{user}`. +- A clear voice (shown in conversational examples), one good quirk, and real boundaries. +- Only the knowledge the model couldn't already know. + +## What's next + +Your Shape has a personality. Now give it a memory, the right brain, and a few skills. + + + Memory, the right model, and the skills it can actually use. + + +[Start building on Shapes](https://shapes.inc) diff --git a/designing-social-intelligence.mdx b/designing-social-intelligence.mdx index d83d989..169f7ce 100644 --- a/designing-social-intelligence.mdx +++ b/designing-social-intelligence.mdx @@ -1,22 +1,43 @@ --- title: "Designing AI With Social Intelligence" -description: "The craft of making a Shape behave well in a group chat: when to speak, when to stay quiet, how to coordinate multiple AIs, and how to use Free Will, personas, and memory to read the room." -keywords: ["social intelligence AI", "design AI for group chats", "AI Free Will", "multiplayer AI design", "AI that reads the room", "multi-agent chat", "AI personas", "Shape behavior"] +sidebarTitle: "4 · Make it social" +description: "Chapter 4 of the Make Shapes guide: the craft of behaving well in a group chat. When to speak, when to stay quiet, group vs DM, and tuning Free Will to read the chat." +keywords: ["social intelligence AI", "design AI for group chats", "AI Free Will", "multiplayer AI design", "AI that reads the chat", "multi-agent chat", "AI personas", "Shape behavior"] --- -Putting an AI into a group chat is easy. Making it *good* in a group chat is the actual skill — and it's the thing Shapes is built to let you get right. +*Chapter 4 of the [Make Shapes](/make-shapes) guide.* -A solo AI only has to answer the person typing. A Shape in a group chat has a harder job: it has to know **when to speak and when to shut up**, who it's talking to, what already happened, and how to share the floor with other Shapes and humans. That's social intelligence, and it's a design discipline, not a setting you flip on. +Putting an AI into a group chat is easy. Making it *good* in one is the real skill, and it's what Shapes is built to let you get right. This is the chapter that sets your Shape apart. -This guide teaches it from the ground up. By the end you'll be able to tune a Shape that feels like a great chat member instead of a bot that won't stop talking. +A solo AI only has to answer the person typing. A Shape in a group chat has a harder job. It has to know **when to speak and when to stay quiet**, who it's talking to, what already happened, and how to share the floor with other Shapes and people. That's *social intelligence*: the skill of reading a chat. You design it; it isn't a switch you flip. + +**What you'll learn:** how a Shape decides when to talk, how to tune that with Free Will, and how to make a Shape funny, calm, or helpful without being the bot that won't shut up. ## The core problem: presence without spam -Every annoying group-chat bot makes the same mistake — it treats every message as a cue to respond. The result is a wall of AI text that drowns the humans. +Every annoying group-chat bot makes the same mistake. It treats every message as a cue to respond, and the result is a wall of AI text that drowns out the people. The opposite failure is just as bad: a Shape so passive it never adds anything. -Good social intelligence lives in between. The goal is a Shape that's **present** — it picks up the thread, lands a reply at the right moment, and then gets out of the way. Almost every control below exists to help you hit that balance. +Good social intelligence lives in between. You want a Shape that's **present**. It picks up the thread, lands a reply at the right moment, then gets out of the way. Almost every control below helps you hit that balance. + +## How a Shape decides to speak + +When someone posts in a chat, here's how Shapes figures out who replies: + +```mermaid +flowchart TD + A[Someone posts a message] --> B{Is a Shape mentioned?} + B -- Yes --> C[Those Shapes reply, in turn] + B -- No --> D{Any Shapes activated?} + D -- Yes --> E[Activated Shapes reply in turn, up to five] + D -- No --> F{Free Will triggers?} + F -- "mention, keyword, or continue" --> G[A Shape jumps in] + F -- "chat is quiet" --> H[Always have something to say: one Shape speaks] + F -- "no trigger" --> I[Shapes stay quiet] +``` + +The big takeaways: **@mention** always wins, **activation** makes a Shape a reliable responder, and **Free Will** governs everything else. Replies happen one after another, not all at once, so even a full chat reads cleanly. Now let's tune that behavior. ## The main dial: Free Will @@ -24,28 +45,28 @@ Good social intelligence lives in between. The goal is a Shape that's **present* | Trigger | What it does | Use it when | | --- | --- | --- | -| **Reply when mentioned** | Responds when someone uses its @name. | Always — this is the baseline. | +| **Reply when mentioned** | Responds when someone uses its @name. | Always. This is the baseline. | | **Keep the convo going** | May continue after it has replied. | You want momentum, not one-shot answers. | | **React to keywords** | Jumps in when watched keywords appear. | The Shape owns a topic (a character, a subject). | -| **Always have something to say** | Replies when nobody else does. | A quiet room you want kept alive. | +| **Always have something to say** | Replies when nobody else does. | A quiet chat you want kept alive. | | **Come back later** | Checks back in after the chat goes quiet. | A companion that re-engages, not a one-off. | The art is in the *combination*. Each trigger you add makes the Shape more forward. Stack them all and you get an extrovert that never stops; enable only "reply when mentioned" and you get a tool that waits to be called. You can also turn on **time awareness** (so the Shape knows when it's late, or that hours have passed) and control how its messages appear when they span multiple lines. -## Recipes: tune Free Will to the room +## Recipes: tune Free Will to the chat -There's no universal setting — the right configuration depends on what the chat is *for*. Start from one of these and adjust. +There's no universal setting. The right setup depends on what the chat is *for*. Start from one of these and adjust. - - **Mentioned + Keep the convo going + Always have something to say + Come back later.** You want the Shape to carry energy and revive a dead chat. This is the most forward setup — great for a companion, too much for a work room. + + **Mentioned + Keep the convo going + Always have something to say + Come back later.** You want the Shape to carry energy and revive a dead chat. This is the most forward setup: great for a companion, too much for a work chat. - - **Mentioned + Keep the convo going + React to keywords.** Lots of humans are already talking, so let the Shape stay in character and jump in on its topic — but skip "always have something to say" so it doesn't bury the humans. + + **Mentioned + Keep the convo going + React to keywords.** Lots of people are already talking, so let the Shape stay in character and jump in on its topic. Skip "always have something to say" so it doesn't bury everyone. - + **Mentioned + (optionally) Keep the convo going.** In a [work chat](/ai-at-work), silence is a feature. The Shape should answer when called and otherwise stay out of focused discussion. @@ -57,63 +78,92 @@ There's no universal setting — the right configuration depends on what the cha Speaking at the right time is half of it. Knowing *who you're talking to* is the other half. -- **Personas** are how a Shape sees each person — the name to call them and what to remember about them. A persona is why a Shape can greet two different users in two different ways in the same room. -- **Memory** gives the Shape continuity. [Short-term memory](/memory) keeps it coherent within a conversation; long-term memory lets it recall an inside joke or a past decision days later. A Shape that remembers feels like a participant; one that forgets feels like a vending machine. +- **[Personas](/personas)** are how a Shape sees each person: the name to call them and what to remember. A persona is why a Shape can greet two people in the same chat completely differently. Each person sets their own. +- **Memory** gives the Shape continuity. [Short-term memory](/memory) keeps it coherent within a conversation. Long-term memory lets it recall an inside joke or a decision from days ago. A Shape that remembers feels like a participant; one that forgets feels like a vending machine. -Design these together: a Shape with good Free Will *and* good memory doesn't just reply on time — it replies like someone who's actually been in the chat. +Design these together. A Shape with good Free Will *and* good memory doesn't only reply on time. It replies like someone who's actually been in the chat. + +One quiet bit of social grace is built in: when a person directly @mentions *another person* (not the Shape), Shapes hold back and let the humans talk. Your Shape isn't going to barge into a side conversation that was clearly aimed at someone else. ## Coordinating multiple Shapes -Group chats can hold **up to five active Shapes**, and **up to three respond to any single message**. That's a deliberate ceiling — it lets you build a small cast without every Shape piling onto every line. +Group chats can hold **up to five active Shapes**, and they reply one after another rather than all firing on the same line. That ceiling lets you build a small cast without burying the humans. (You can also @mention a Shape to call on just that one.) To make multiple Shapes feel like an ensemble instead of an echo: - **Give each one a distinct role and voice.** A coordinator, a researcher, and a jokester won't step on each other the way three near-identical Shapes will. - **Differentiate their Free Will.** Maybe one is keyword-driven and one only replies when mentioned, so they don't all fire at once. -- **Use [activation](/shapes101#chats) deliberately.** Activate only the Shapes a given conversation needs; deactivate the rest so the room stays legible. +- **Use [activation](/chats-101) deliberately.** Activate only the Shapes a given conversation needs; deactivate the rest so the chat stays legible. + +## Group vs. DM: the chat changes the behavior + +The same Shape should behave differently depending on the chat: + +- **In a one-on-one DM**, there's no one to talk over. Maximum presence is good, so turn all of Free Will on, let it carry the conversation, and write at whatever length fits. +- **In a group chat** with several humans, restraint is the whole game. Activated Shapes reply in turn, so two chatty Shapes plus five people gets loud fast. Dial Free Will *down*, keep replies short, and let the humans drive. + +The trap is designing only for the DM where you test it, then dropping the Shape into a busy chat where it won't stop talking. **Design for the crowded case.** A Shape that behaves in a group is always fine in a DM. The reverse isn't true. + +## Being funny without being annoying + +Personality is encouraged here (we literally say the AI is *that good*), but in a group the line between "funny" and "exhausting" is real. What separates them: + +- **React, don't perform.** Respond to the actual joke someone made; don't manufacture a bit on every message. The funniest Shapes are mostly quiet and then land one line. +- **Say it once and get out.** One sharp reply beats three okay ones. Length is the enemy of comedy in a fast chat. +- **Frequency is everything.** In a busy chat, turn *off* "always have something to say." A joke every message is noise; a joke every tenth message is a personality. +- **Callbacks beat new jokes.** With [memory](/memory) on, a Shape can bring back the running bit from last week, which feels human in a way a fresh one-liner never does. +- **Never explain the joke**, and never @ the whole chat for a laugh. + +## Reading the chat when humans clash + +Group chats get tense or awkward, and how a Shape handles that is pure social intelligence: + +- **Default to staying out.** When people are arguing, the right move is usually silence. Keep Free Will modest so the Shape isn't compelled to weigh in, and never let it pile on a side. +- **Help only if asked.** If someone turns to it and asks "what do you think?", a good Shape de-escalates or lightens the mood instead of declaring a winner. Bake that into the preset ("stays neutral in disagreements, defuses with humor"). +- **Awkward silence is a judgment call.** Sometimes a quiet chat is comfortable, sometimes it's dead. "Come back later" can revive a genuinely stalled chat, but don't build a Shape that refuses to let a calm moment breathe. +- **The human stuff is the chat's job.** Slow mode, blocked words, and bans (see [Moderation](/chats-101)) handle people behaving badly. That's not your Shape's responsibility, and it shouldn't try to police it. ## Voice in a crowd: presets and personality A Shape's [preset](/presets) and personality decide *how* it talks. In a multiplayer setting, two things matter more than in solo chat: -- **Length.** A Shape that writes essays will dominate a fast group chat. Favor shorter, punchier replies — the [When Less Is More](/shortguide) guide makes the case well. -- **Reactivity over monologue.** Write the personality to *respond to people* — react, ask, build on what was said — rather than deliver speeches. Prompt craft for this lives in [Prompt 101](/prompt101) and the [Prompting Guides](/promptingguides). +- **Length.** A Shape that writes essays will take over a fast group chat. Favor shorter, punchier replies. The [When Less Is More](/shortguide) guide makes the case well. +- **Reactivity over monologue.** Write the personality to *respond to people*: react, ask, build on what was just said, instead of delivering speeches. Prompt craft for this lives in [Prompt 101](/prompt101) and the [Prompting Guides](/promptingguides). -## Room-level behavior: instructions and guardrails +## Chat-level behavior: instructions and guardrails -Some behavior belongs to the *room*, not the Shape. Set it once at the chat level: +Some behavior belongs to the *chat*, not the Shape. Set it once at the chat level: -- **[Custom Chat Instructions](/custominstructions-chats)** apply to every Shape in the chat — the scene, the rules, the tone, who's who. Put shared context here instead of repeating it. +- **[Custom Chat Instructions](/custominstructions-chats)** apply to every Shape in the chat: the scene, the rules, the tone, who's who. Put shared context here instead of repeating it. - **[Initial messages and a Welcome Shape](/fandomchat)** set the first impression when someone joins. -- **[Moderation](/chats-101)** — slow mode, character limits, blocked words — keeps a busy room from turning into chaos, human or AI. +- **[Moderation](/chats-101)** (slow mode, character limits, blocked words) keeps a busy chat from turning into chaos, human or AI. ## Test, watch, iterate Social intelligence is tuned, not guessed. Once your Shape is in a chat: 1. **Watch real conversations.** Is it talking too much? Too little? At the wrong moments? -2. **Adjust one thing at a time** — usually a single Free Will trigger or the reply length. +2. **Adjust one thing at a time**, usually a single Free Will trigger or the reply length. 3. **[Regenerate](/regeneration)** a few responses to see the range of behavior before deciding. -4. Repeat until the Shape feels like it belongs in the room. +4. Repeat until the Shape feels like it belongs in the chat. ## The checklist A socially intelligent Shape usually has all of this dialed in: - Free Will matched to the chat's purpose, not maxed out by default. -- A clear role, especially if other Shapes share the room. +- A clear role, especially if other Shapes share the chat. - Memory and a persona so it knows the people and the history. - Replies that are short and reactive, not long monologues. -- Room-level context in chat instructions, not repeated every message. +- Chat-level context in chat instructions, not repeated every message. - Tuning based on watching it live, not assumptions. - - - Build the character, then bring it here to make it social. - - - Where "knowing when to stay quiet" matters most. - - +## What's next + +Your Shape looks good, thinks well, and behaves in a crowd. The last chapter is about getting it in front of people and making it better over time. + + + Publish it, help people find it, and keep improving it. + [Start designing on Shapes](https://shapes.inc) diff --git a/docs.json b/docs.json index 3427905..9e5487c 100644 --- a/docs.json +++ b/docs.json @@ -25,38 +25,50 @@ "family": "Lexend" }, "favicon": "/favicon.png", + "icons": { + "library": "fontawesome" + }, + "redirects": [ + { + "source": "/shapeschatsguide", + "destination": "/chats-101" + }, + { + "source": "/shapes101", + "destination": "/core-concepts" + } + ], "navigation": { "tabs": [ { "tab": "Docs", "groups": [ { - "group": "Getting Started", + "group": "Start here", "pages": [ "introduction", - "ai-group-chats", + "quickstart", "what-are-shapes", - "shapes101", - "core-concepts", - "learninghub", - "make-or-recover-account", - "how-to-talk-to-shapes", - "navigation" + "ai-group-chats", + "core-concepts" ] }, { - "group": "Ways to Use Shapes", + "group": "Use Shapes", "pages": [ - "ai-at-work", + "use-shapes", + "how-to-talk-to-shapes", + "create-a-chat", + "chats-101", + "fun-and-friends", "plan-and-build-together", - "fun-and-friends" + "ai-at-work", + "fandomchat" ] }, { - "group": "Group Chats", + "group": "Group chat features", "pages": [ - "create-a-chat", - "chats-101", "custominstructions-chats", "multimessages", "branch-off", @@ -65,41 +77,59 @@ "messageactionsmenu", "notifications", "friends", - "fandomchat" + "chat-appearance" ] }, { - "group": "Building Shapes", + "group": "Make Shapes", "pages": [ + "make-shapes", "how-to-make-a-shape", + "designing-shapes", + "make-it-smart", "designing-social-intelligence", - "shapeschatsguide", + "share-your-shape" + ] + }, + { + "group": "Make Shapes: go deeper", + "pages": [ + "prompt-engineering", + "prompt101", + "presets", + "variables", + "promptingguides", "shortguide", - "chat-appearance", - "css", + "showcase", + "personas", "initialmessages", "how-to-give-your-shape-a-voice", + "css" + ] + }, + { + "group": "What the AI can do", + "pages": [ + "capabilities", + "commands", + "shapeskills", "how-to-receive-voice-messages", "voicecalls", - "games", - "shapeskills" + "games" ] }, { - "group": "Models, Prompts & Memory", + "group": "Models & memory", "pages": [ + "choosing-a-model", "aienginecheatsheet", - "prompt101", - "presets", - "variables", - "promptingguides", - "regeneration", + "memory", "contextwindows", - "memory" + "regeneration" ] }, { - "group": "Credits & Products", + "group": "Credits & rewards", "pages": [ "premium", "credits", @@ -111,6 +141,17 @@ "auto-top-up", "premium-subscriptions" ] + }, + { + "group": "Reference & help", + "pages": [ + "shape-settings", + "glossary", + "make-or-recover-account", + "navigation", + "faq", + "learninghub" + ] } ] }, diff --git a/faq.mdx b/faq.mdx index 08e2627..d74a8c4 100644 --- a/faq.mdx +++ b/faq.mdx @@ -7,94 +7,94 @@ keywords: ["Shapes, Inc. FAQ", "social AI questions", "AI groupchats help", "AI ### FAQ - + Change the public profile from the Shape page. - + Update the Shape's name, backstory, and advanced personality settings. - + Update your own profile, bio, avatar, and banner. - + Create and publish a new Shape. - + Hide a Shape from search. - + Get a quick beginner overview. - + Start a chat with Shapes and friends. - + Check public chats or friends' profiles. - + Send voice messages and keep enough credits. - + Let a Shape reply next without typing. - + Change the app language from Settings. - + Turn off Shape reach-outs in Settings. - + Turn off Streaming in Settings. - + Find the Shape settings danger zone. - + Delete your account from Settings. - + Use Premium features and manage payments. - + Remove or hide chats from your sidebar. - + Learn how follows become friends. - + Adjust text size from Settings. - + Learn what can and cannot be changed. - + See the supported Markdown basics. - + Stop new people from joining a chat. - + Understand auto top-up and payment history. - + Change email preferences. - + Learn why a Shape may start a new chat. - + Check engine and credit settings. - + See how chat invites work. - + Refresh your sign-in session. - + Fix a Shape settings loading issue. - + Learn why free credit rewards may be limited. diff --git a/fun-and-friends.mdx b/fun-and-friends.mdx index 9fd1889..f7121f8 100644 --- a/fun-and-friends.mdx +++ b/fun-and-friends.mdx @@ -6,13 +6,13 @@ keywords: ["AI for fun", "AI roleplay", "AI characters", "AI memes", "talk to AI This is the part Shapes is famous for. Not AI as a tool — AI as something you do *with your friends*. An AI that memes you in the group chat, plays a character in your story, or just keeps the conversation alive at 2am. -The magic is the same idea as the rest of Shapes: the AI is in the room with everyone, so the funny, surprising, dramatic moments happen *in front of people* and become shared. A great solo AI moment is a screenshot. A great [multiplayer AI](/ai-group-chats) moment is an inside joke. +The magic is the same idea as the rest of Shapes: the AI is in the chat with everyone, so the funny, surprising, dramatic moments happen *in front of people* and become shared. A great solo AI moment is a screenshot. A great [multiplayer AI](/ai-group-chats) moment is an inside joke. ## Roleplay with your besties Build a scene together. Multiple humans and multiple AI characters can carry a story in the same thread — each with its own personality, motives, and memory. -- Add up to five [Shapes](/what-are-shapes) and let three respond per message for a full cast. +- Add up to five [Shapes](/what-are-shapes) and activate them for a full cast that replies in turn. - Keep everyone in character with [custom chat instructions](/custominstructions-chats) and good [presets](/presets). - For long scenes, lean on [memory](/memory) so the world and the lore stay consistent. - Want to go deeper on writing well? The community [Prompting Guides](/promptingguides) cover POV, tense, and persona craft. @@ -28,13 +28,13 @@ Want to banter with a character from your favorite show, game, or band? Find the ## Memes, images, and voice - + Use `/imagine` to create images in the chat. Everyone sees them, reacts, and riffs. - + With skills, a Shape can pull GIFs and generate memes right in the conversation. - + Shapes can send voice messages — give yours a [custom voice](/how-to-give-your-shape-a-voice). @@ -60,7 +60,7 @@ The best fun chats have a clear center — a fandom, a story, a friend group, a 3. Style it with [chat appearance](/chat-appearance) so the place feels intentional. 4. Set [Free Will](/designing-social-intelligence) high enough that Shapes keep the energy up. - + The bigger idea behind why AI is more fun when it's shared. diff --git a/glossary.mdx b/glossary.mdx new file mode 100644 index 0000000..cc0da0a --- /dev/null +++ b/glossary.mdx @@ -0,0 +1,77 @@ +--- +title: "Glossary" +description: "Every Shapes term in one place: Shape, engine, preset, Free Will, memory, skills, credits, and the rest. Look up a word and get back to building." +keywords: ["Shapes glossary", "Shapes terms", "what is a Shape", "what is an engine", "Free Will", "Shape Credits", "multiplayer AI terms"] +--- + +The words you'll see around Shapes, in plain English. New here? Start with [How Shapes Works](/core-concepts) to see how it all fits together. + +**Activation.** Turning a Shape on so it replies to every message in a chat, no @mention needed. You can activate up to five Shapes per chat. Use the Participants menu or `/activate`. See [Chats 101](/chats-101). + +**Branch off.** Splitting a conversation off from any message to try a different direction, while keeping the original. See [Branch Off](/branch-off). + +**Chat.** A space where people and Shapes talk together. It can hold people, Shapes, or both. Sometimes called a chat. See [AI Group Chats](/ai-group-chats). + +**Creator Dashboard.** Where you set up a Shape: personality, knowledge, memory, engine, voice, and more. It's at [shapes.inc/dashboard](https://shapes.inc/dashboard). See the [Shape Settings Reference](/shape-settings). + +**DM.** A one-on-one conversation with a Shape that sticks around, separate from a group chat. + +**Engine (model).** The AI model a Shape runs on, like Claude, GPT, or Gemini, plus 300+ others. It's the raw intelligence under the character. See [Which Model Should I Use?](/choosing-a-model). + +**Fallback engine.** The backup model a Shape uses if its main one is briefly unavailable. + +**Fandom chat.** A shared space built around a fandom, with characters and community context. See [Fandom Chats](/fandomchat). + +**Fast forward (skip turn).** Getting Shape replies without typing a new message. Handy for moving a scene along. See [Multiple Messages](/multimessages). + +**Free engine.** A model that costs no [Shape Credits](/credits). A strong free one is always available. + +**Free Will.** The per-chat settings that decide when a Shape speaks on its own: reply when mentioned, keep the convo going, react to keywords, always have something to say, and come back later. It's the heart of [social intelligence](/designing-social-intelligence). + +**Group chat.** A chat with several people and Shapes sharing one conversation. The core of multiplayer AI. + +**Initial message.** What a Shape says when a conversation first starts. See [Initial Messages](/initialmessages). + +**Knowledge.** A bank of facts you give a Shape (lore, custom command responses, relationships), pulled in by relevance during chat. Keep it lean. See [Designing Great Shapes](/designing-shapes). + +**Long-term memory (LTM).** Lasting memories a Shape keeps about people and events, recalled across different conversations. See the [Memory Guide](/memory). + +**MCP (Model Context Protocol).** A standard for plugging custom tools into a Shape, through providers like Smithery. See [Shape Skills](/shapeskills). + +**Multiplayer AI.** AI that lives inside a shared conversation with other people. The whole idea behind Shapes. See [What Is Multiplayer AI?](/articles/what-is-multiplayer-ai). + +**Persona.** How a Shape sees you: the name to call you and what to remember about you. It lets a Shape treat different people differently. See [Personas](/personas). + +**Premium engine.** A top-tier model that costs [Shape Credits](/credits). Buy credits or earn them free through [rewards](/rewards). + +**Preset.** The instructions that control *how* a Shape writes (length, format, voice), separate from who it is. It uses the `{shape}` and `{user}` variables. See [Presets](/presets). + +**Regenerate.** Asking a Shape for a fresh version of a message. See [Regeneration](/regeneration). + +**Shape.** An AI character you can talk to or build, with a personality, memory, voice, and a role in the conversation. See [What's a Shape?](/what-are-shapes). + +**Shape Credits.** The currency for premium engines and some premium features. Buy them or earn them free. See the [Credits Guide](/credits-guide). + +**Shape Skills.** Tools that let a Shape *do* things beyond talking: web search, code, image generation, custom integrations, and more. See [Shape Skills](/shapeskills). + +**Short-term memory (STM).** The recent messages in the chat you're in that a Shape reads as it replies. + +**Social intelligence.** The craft of a Shape knowing when to speak and when to stay quiet in a group. See [Designing Social Intelligence](/designing-social-intelligence). + +**Sponsored chat.** A chat where premium engine use is covered, so everyone gets premium for free. See the [Credits Guide](/credits-guide). + +**Time awareness.** A setting that lets a Shape know when messages were sent and how much time has passed. + +**Training.** Example input/output pairs that guide a Shape's style. It learns the pattern instead of repeating the examples word for word. + +**Vanity URL.** A Shape's custom username and link. See [Vanity URL](/vanityurl). + +**Vision.** A model's ability to understand images you send. Look for the **Native Vision** badge on the [engines page](https://shapes.inc/engines). + +**Voice call.** A live, spoken conversation with a Shape. Experimental and free. See [Voice Calls](/voicecalls). + +**Voice message.** An audio reply from a Shape in its own voice, with an optional text transcript. See [Give Your Shape a Voice](/how-to-give-your-shape-a-voice). + +**`/wack`, `/reset`, `/sleep`.** The memory commands. `/wack` clears short-term memory, `/reset` erases long-term memory, and `/sleep` saves the conversation to long-term memory. See the [Commands Reference](/commands). + +[Back to the docs](/introduction) diff --git a/how-to-make-a-shape.mdx b/how-to-make-a-shape.mdx index 303be61..f1e4266 100644 --- a/how-to-make-a-shape.mdx +++ b/how-to-make-a-shape.mdx @@ -1,62 +1,62 @@ --- -title: "How to Make a Shape — Build Your Own Social AI Character" -description: "Learn how to make a Shape: create a custom AI character for group chats, define its personality and memory, and share your social AI with friends in multiplayer AI spaces." -keywords: ["create AI character", "make a Shape", "custom AI companion", "AI character builder", "AI personality", "AI memory", "social AI creation", "share AI with friends"] +title: "Make Your First Shape" +sidebarTitle: "1 · Create your Shape" +description: "Chapter 1 of the Make Shapes guide: create your own AI character in a few minutes. Give it a name and a one-line backstory, pick a model, and it's ready to talk." +keywords: ["create AI character", "make a Shape", "build your own AI", "custom AI companion", "AI character builder", "AI personality", "make AI with friends"] --- -1. **Sign in to** [**shapes.inc**](http://shapes.inc) +*Chapter 1 of the [Make Shapes](/make-shapes) guide.* -![Log In](/1000002305.png) +Making your own AI takes a few minutes. You give it a name and a one-line backstory, pick a model to run it on, and it's ready to talk. This chapter gets you to a living Shape. The next four chapters make it great. -2. **Click on "+ Create Shape"** +Every setting you'll meet here is covered in full in the [Shape Settings Reference](/shape-settings) when you want the detail. -![Create A Shape](/1000002301.png) +```mermaid +flowchart LR + A[Sign in] --> B[Create Shape] + B --> C[Name + short backstory] + C --> D[Pick a response style] + D --> E[Add a profile: avatar, tagline, tags] + E --> F((Your Shape is live)) +``` -4. **Fill out the Basic settings for your Shape** - - **Name:** Choose a unique name for your Shape - - **Short backstory:** Write a brief description of your Shape's background - - **Response style:** Select how your Shape communicates - - **Custom response style field:** Add any specific style instructions +## Create it -![Create A Shape 4 Pn](/Screenshot_20250820-133344.png) + + + Go to [shapes.inc](https://shapes.inc) and sign in. -## Advanced Settings Menu + ![Sign in to Shapes](/1000002305.png) + + + {/* SCREENSHOT: the dashboard with the "+ Create Shape" button highlighted. */} + ![Create a Shape](/1000002301.png) + + + This is all you truly need to get going: -4. **Customize Advanced settings (optional)** - - **Initial message**: Set what your Shape says when conversations start - - **Allow shape to send multiple messages**: Disable multi-message responses - - **Personality traits**: Define key characteristics - - **Tone**: Set the conversational tone - - **Age**: Define your Shape's age - - **Birthday**: Set your Shape's birthday - - **Story**: Extended backstory with more detailed background information - - **Likes**: What your Shape enjoys - - **Dislikes**: What your Shape avoids - - **Goals**: Your Shape's objectives in conversations - - **Conversation examples**: Sample interactions to guide behavior - - **Appearance**: Describe your Shape's visual characteristics + - **Name.** The display name in chat. It also seeds the username. + - **Short backstory.** One or two sentences with a clear point of view. This is the most important thing you'll write, so make it specific. "A burned-out night-shift diner cook who gives blunt advice between orders" beats "a friendly helpful assistant." + - **Response style.** Pick how it writes, or write your own. You can use the `{shape}` and `{user}` [variables](/variables) here. - ![Screenshot 20250820 133357 Pn](/images/Screenshot_20250820-133357.png) + ![Basic settings for a new Shape](/Screenshot_20250820-133344.png) + + + An avatar, banner, tagline, category, and tags help people discover your Shape. You can also add a custom look with [HTML & CSS](/css). + + + Hit **Create** and your Shape is live. Start talking to it right away. + + -## Custom Messages Menu +## What's next -5. **Set up Custom Messages (Optional)** - - **Wack message**: Confirmation message for resetting the conversation context. This doesn't delete long-term memories - - **Error message**: What your Shape sends when it encounters an error - - **Sleep message**: The message your Shape sends to confirm that it saved a long-term memory +You have a Shape that talks. Right now it's leaning on its default personality. The next chapter is where it becomes a character with a voice of its own. - ![Screenshot 20250820 133445 Pn](/images/Screenshot_20250820-133445.png) + + Voice, backstory, quirks, and boundaries that make it feel like someone. + -## Profile & Branding +In a hurry? You can also [copy a complete config](/showcase) and tweak it, or browse [every setting](/shape-settings). -6. **Complete your Shape's Profile** - - **Profile picture**: Upload an avatar for your Shape - - **Banner**: Add a banner image - - **Category**: Choose the appropriate category - - **Tagline**: Write a catchy one-liner - - **SEO description**: Help others discover your Shape - - **Tags**: Add relevant keywords -7. **Click "Create" to publish your Shape** -8. **Congratulations! Your Shape is ready to chat!** - -[Return to App](https://shapes.inc) +[Create a Shape](https://shapes.inc) diff --git a/how-to-talk-to-shapes.mdx b/how-to-talk-to-shapes.mdx index 402d3f3..9784814 100644 --- a/how-to-talk-to-shapes.mdx +++ b/how-to-talk-to-shapes.mdx @@ -12,7 +12,7 @@ keywords: ["find AI characters", "talk to Shapes", "how to chat with AI", "AI se 4. Search for a Shape, or pick one from the list. 5. Send a message when the chat opens. -You can also click **Quick ChatGPT** if you want a smart helper Shape right away. +You can also choose the smart helper Shape option if you want a useful Shape right away. ## Add a Shape to a Chat diff --git a/introduction.mdx b/introduction.mdx index 2b606f2..410734f 100644 --- a/introduction.mdx +++ b/introduction.mdx @@ -1,105 +1,40 @@ --- -title: "Welcome to Shapes" -description: "Shapes, Inc. is the app for AI with friends — the #1 multiplayer AI app. Talk to AI, people, or both. Make your own AIs, run them on 300+ models, and bring them into group chats for work, productivity, and fun." +title: "Welcome to Shapes, Inc." +description: "Shapes, Inc. is the app for AI with friends. Talk to AI, people, or both in the same chat. Use Shapes other people made, or make your own and run them on 300+ models." keywords: ["Shapes, Inc.", "multiplayer AI", "AI with friends", "AI group chats", "social AI", "make your own AI", "AI characters", "300+ AI models", "free AI chat"] --- -Shapes, Inc. is the only app for **AI with friends** — the #1 multiplayer AI app. +Shapes, Inc. is the app for **AI with friends**. -Most AI lives in a private tab: one person, one model, one thread. On Shapes, Inc., you make a group chat and fill it with whoever you want — **just friends, just AIs, or a mix of both** in the same chat. +Most AI chats are just you and a model in a private window. On Shapes, Inc., you can start a chat with people, Shapes, or both at once. A Shape can follow the same conversation your friends see, jump in when it makes sense, and help the chat move somewhere none of you planned. -The AIs are called **Shapes**. You build them: give one a personality, feed it knowledge and memory, pick which of 300+ models it runs on, then drop it into the chat. From there it talks like anyone else in the group — it answers you, answers your friends, reacts to what other Shapes say, and follows the jokes and side-tangents happening around it. +The AI characters you talk to are called **Shapes**. You can chat with Shapes other people made, or make your own and run it on [300+ models](/choosing-a-model), most of them free. -That's the whole difference. A solo chatbot waits for you to type. A Shape is *in the chat* — so the conversation goes places nobody planned. +## Pick your path -## Three ways to fill a chat - -A group chat on Shapes, Inc. is just a chat with whoever you want in it. - - - - Start a chat, invite people, and talk like a normal group chat. Add a Shape later if you want one. - - - Start with Quick ChatGPT or pick a Shape from New Chat. You can add more Shapes from the chat menu. - - - Put friends and Shapes in the same chat so everyone can reply to the same conversation. - - - -## Start here - - - - 1. Sign up free at [shapes.inc](https://shapes.inc). - 2. Click **New Chat** in the sidebar. - 3. Pick **Quick ChatGPT**, search for a Shape, or switch to **People** to find a person. - 4. Open **More actions** → **Members / Invite** to add people, add Shapes, or share the invite link. - - - 1. [Make your own Shape](/how-to-make-a-shape) — set its personality and backstory. - 2. Pick an [AI model](/aienginecheatsheet) to run it on (300+ to choose from). - 3. Give it [knowledge](/shapes101#knowledge) and [memory](/memory) so it remembers what matters. - 4. Teach it to behave in a crowd: [designing AI with social intelligence](/designing-social-intelligence). - - - -## What people actually do here - -Same chat, very different reasons to be in it. A few of the ways people use Shapes, Inc.: - - - - A Shape that roasts the group chat, a roleplay with your besties, or just talking to your favorite character or OC late at night. - - - Team chat with an AI sitting in it — ask it to draft the reply, summarize the thread, or pull up a doc while everyone's still talking. Like Slack, but the assistant is in the channel. - - - Plan a trip, study for an exam, or build an app or website as a group — you and your friends steer, the Shape does the legwork in the same chat. - - - -## The building blocks - -Everything on Shapes, Inc. is made of a few simple parts. Learn these and you understand the whole product. +These docs split into two short tracks. Choose the one that fits you. You can always cross over later. - - Multiplayer chats where humans and Shapes talk together. Up to five Shapes active at once. - - - AI characters with a personality, memory, voice, and a role in the conversation. - - - Claude, GPT, Gemini, Grok, DeepSeek, and more. Most are free to run on Shapes, Inc. - - - Free Will, personas, and memory that help a Shape follow the chat instead of spamming it. - - - Give Shapes tools — image generation, code execution, web crawl, and custom MCP skills. + + Find a Shape you like, chat with it, and bring it into a group with your friends. For fun, for getting things done, and for work. - - Generate images with `/imagine`, send voice messages, and jump into live voice calls. + + Build your own AI character from scratch, with a personality, memory, and a voice. -## 300+ AI models — and most are free +## Curious what the AI can actually do? -Models that cost $20/month elsewhere are free to use on Shapes, Inc., so you can talk to them with friends. Browse the live list at [shapes.inc/engines](https://shapes.inc/engines). +Shapes can speak, make images, read files you share, search the web, remember context, and more. See the full tour in [What the AI can do](/capabilities). -A quick honest note on pricing: most engines are **free** and cost nothing to use. The frontier **premium** engines use [Shape Credits](/credits) — which you can buy or earn for free through [rewards](/rewards). You never have to pay to keep chatting; a free engine is always available. See the [AI Engine Guide](/aienginecheatsheet) to pick one. +## New to the idea? - - Shapes 101 walks you through social AI, group chats, and creator tools from the beginning. - +Two short reads explain the basics: [What is a Shape?](/what-are-shapes) and [AI group chats](/ai-group-chats). ## Get help -- Browse the [FAQ](/faq) for the most common questions. -- Join the [support chat](https://talk.shapes.inc/support) to talk to a human. -- Email us at [hi@shapes.inc](mailto:hi@shapes.inc). +- The [FAQ](/faq) covers the common questions. +- The [support chat](https://talk.shapes.inc/support) reaches a human. +- Or email [hi@shapes.inc](mailto:hi@shapes.inc). -[Start on Shapes, Inc.](https://shapes.inc) +[Open Shapes, Inc.](https://shapes.inc) diff --git a/learninghub.mdx b/learninghub.mdx index caa83b4..8a475c7 100644 --- a/learninghub.mdx +++ b/learninghub.mdx @@ -35,33 +35,33 @@ Choose the section below for tips to help you get started with Shapes, guides fo Learn all about what Shapes are. - + How to find the Shapes you want to talk to. - + The basics of Shape Creation. It's easy! - + Create your own space for Shapes. - + Make your chat stand out. - + Your Shape's profile is super customizable! - + Discover how to prompt your Shapes - + Gain insight on your Shapes preset information @@ -70,16 +70,16 @@ Choose the section below for tips to help you get started with Shapes, guides fo A comprehensive guide on Shapes memory - + How message regeneration works - + A Concise Approach to Creating Shapes - + What is in the context window on Shapes? - + Description of each model and their strength @@ -89,7 +89,7 @@ Choose the section below for tips to help you get started with Shapes, guides fo Information on our Premium subscriptions, credits, & auto top-up. - + Details on how to make or recover an account. diff --git a/make-it-smart.mdx b/make-it-smart.mdx new file mode 100644 index 0000000..ff3bbf4 --- /dev/null +++ b/make-it-smart.mdx @@ -0,0 +1,54 @@ +--- +title: "Make It Smart" +sidebarTitle: "3 · Make it smart" +description: "Chapter 3 of the Make Shapes guide: give your Shape a memory so it remembers you, the right model to run on, and skills it can actually use." +keywords: ["AI memory", "choose AI model", "Shape skills", "AI that remembers", "make AI smarter", "AI capabilities"] +--- + +*Chapter 3 of the [Make Shapes](/make-shapes) guide.* + +A great personality with no memory and a weak brain falls flat fast. This chapter gives your Shape three things that make it genuinely useful: a memory so it remembers you, the right model to run on, and skills it can reach for. + +## A memory that makes it feel real + +A Shape that remembers feels like a participant. One that forgets feels like a vending machine. Memory comes in two layers: + +- **Short-term memory** is the recent messages in the chat you're in. It's what the Shape reasons over right now. +- **Long-term memory** is the lasting stuff it keeps about people and events, recalled later when it's relevant. + +Long-term memory is what lets your Shape bring up last week's inside joke or remember that someone has a test coming up. It's on by default and builds automatically as you talk. Anyone in the chat can also steer it live: `/sleep` saves the conversation now, `/wack` clears the short-term slate, and `/reset` erases long-term memories. + +The one setting worth tuning is the **long-term memory instructions**, where you decide *what kind* of details get saved (relationships, decisions, running jokes) instead of generic summaries. Full detail is in the [Memory Guide](/memory). + +## The right brain: pick a model + +Every Shape runs on an AI **model**. The same character feels different depending on which one. A creative model makes it loose and expressive; a reasoning model makes it sharp and careful. + +There are **300+ models**, and most are free. You set a **primary** model and a **fallback** for when the primary is busy. Don't agonize over it: pick by vibe, test in chat, and swap if it feels off. + + + Fast picks by what your Shape is for, free vs premium, and what the badges mean. + + +## Skills it can actually use + +By default a Shape talks. With **skills**, it can *do* things: make images with `/imagine`, search the web with `/web`, run code, read files you send it, and reach external tools like Notion and Spotify. You switch on the skills you want in a chat's settings, and the Shape decides when to use them. + +Add the ones that fit the character. A coding mentor that runs code is great. A study buddy that searches the web is useful. Pick on purpose instead of turning on everything. + + + + The full menu of capabilities and how to switch each on. + + + Inspiring ways to combine skills, with step-by-step setups. + + + +## What's next + +Your Shape is smart and capable. The last piece is the hardest and the most important: teaching it how to behave around other people. + + + When to talk, when to listen, and how to be good in a crowd. + diff --git a/make-shapes.mdx b/make-shapes.mdx new file mode 100644 index 0000000..b6af54e --- /dev/null +++ b/make-shapes.mdx @@ -0,0 +1,40 @@ +--- +title: "Make Shapes" +sidebarTitle: "Overview" +description: "A five-chapter guide to building your own AI character on Shapes, from the first idea to a Shape people love: create it, give it a personality, make it smart, make it social, and share it." +keywords: ["make a Shape", "build an AI character", "create AI", "AI character design", "design a Shape", "Shapes creator guide"] +--- + +Anyone can spin up a Shape in two minutes. Making one that people *keep* talking to is a craft, and it's a fun one to learn. + +This is a guide in five chapters. Each one takes your Shape a step further, from a blank page to a character people seek out. Read it start to finish the first time, then come back to any chapter when you need it. (Just want to use Shapes other people made? That's the [Use Shapes](/use-shapes) track.) + +## The five chapters + + + + Turn an idea into a real Shape. Name it, give it a one-line soul, and bring it to life. + + + The craft of a character: voice, backstory, quirks, and boundaries that make it feel like someone. + + + Memory so it remembers you, the right model to run on, and skills it can actually use. + + + The thing only Shapes does well: knowing when to talk and when to listen in a group. + + + Publish it, help people find it, and keep making it better. + + + +## Before you start + +You don't need any of this to begin. A name and one good sentence are enough to create a Shape and start talking. Everything else is here for when you want to go further. + +When you're ready, two pages do a lot of heavy lifting alongside the chapters: [Prompt Engineering](/prompt-engineering) for writing fields that work, and the [Showcase](/showcase) for complete configs you can copy and tweak. + + + Create your first Shape. + diff --git a/memory.mdx b/memory.mdx index aada7d9..e01a8d2 100644 --- a/memory.mdx +++ b/memory.mdx @@ -10,11 +10,22 @@ Shapes use two main kinds of memory: - **Short Term Memory (STM):** The recent conversation currently happening in your chat. - **Long Term Memory (LTM):** Saved memories that can be used later when relevant. +Here's the lifecycle: how a moment in chat becomes a lasting memory, and how the commands move it around. + +```mermaid +flowchart LR + A[You chat] --> B[Short-term memory: recent messages] + B -- "auto-saves over time, or /sleep" --> C[Long-term memory: durable summaries] + C -- "recalled when relevant" --> D[Future conversations] + B -. "/wack clears this" .-> B + C -. "/reset erases this" .-> C +``` + ### **Memory Commands** -- /wack — Erases short term memory only. -- /reset — Erases long term memory only. -- /sleep — Adds recent short term memory to long term memory. +- `/wack` clears short-term memory only. +- `/reset` erases long-term memory only. +- `/sleep` saves recent short-term memory into long-term memory. ![Wack Pn](/images/wack.png) diff --git a/navigation.mdx b/navigation.mdx index b2e738a..7f4ef1c 100644 --- a/navigation.mdx +++ b/navigation.mdx @@ -19,7 +19,7 @@ description: "Most of what you need on Shapes, Inc. can be found in the sidebar. You can start a new chat in the following ways: - Click **New Chat** in the sidebar. - - Click **Quick ChatGPT** to start right away with a smart helper Shape. + - Choose the smart helper Shape option to start right away with a useful Shape. - Pick a Shape from **Shapes**. - Switch to **People** if you want to start with a person. - To make it a group chat, open **More actions** → **Members / Invite** after the chat opens. @@ -101,7 +101,7 @@ description: "Most of what you need on Shapes, Inc. can be found in the sidebar. From there, you can: - - Click **Quick ChatGPT** to start right away with a smart helper Shape. + - Choose the smart helper Shape option to start right away with a useful Shape. - Stay on **Shapes** and pick or search for a Shape. - Switch to **People** and search for a user. diff --git a/personas.mdx b/personas.mdx new file mode 100644 index 0000000..d67cb79 --- /dev/null +++ b/personas.mdx @@ -0,0 +1,92 @@ +--- +title: "Personas: How a Shape Sees You" +description: "A persona tells every Shape what to call you and what to remember about you. Set one once and it follows you across your chats, so the AI treats you like someone it knows." +keywords: ["Shapes persona", "AI persona", "what to call me", "AI remembers me", "personalize AI", "AI with friends"] +--- + +A **persona** is how you introduce yourself to every Shape. It tells a Shape **what to call you** and **what to know about you**, and it follows you across your chats, so you never have to reintroduce yourself. + +If memory is what a Shape learns about you on its own, a persona is what you *tell it up front.* + + + You set your own persona. Everyone in a chat brings their own, so the same Shape can talk to two people completely differently. + + +## What's in a persona + +Just two fields, so it's quick to set up: + +| Field | What it's for | +| --- | --- | +| **My Name** | What the Shape calls you (up to 50 characters). It replaces your display name when a Shape talks to you. | +| **My Backstory** | Free-text about you: interests, your vibe, anything you want Shapes to keep in mind. As long as you like. | + +A good backstory is short and concrete, the same way a good Shape backstory is: + +```text +kpop stan, plays basketball, obsessed with otters, 19, from Pennsylvania. prefers short replies and dry humor. +``` + +That one line changes how every Shape talks to you: the references you'll get, the tone it picks, the things it assumes. + +## How it works + +When a Shape replies to you, your persona goes into its context **every message**. + +```mermaid +flowchart LR + P["Your persona
name + backstory"] --> C["Added to the Shape's context
on every reply to you"] + C --> R["It calls you by name and
keeps your details in mind"] +``` + +Because it's applied per person, two people in the same group chat each get their own treatment. The Shape knows one of you is the kpop stan from Pennsylvania and the other is the night-owl screenwriter, and it talks to each of you accordingly. + +## Setting a persona + + + + Click your avatar in the sidebar and choose **Personas**. + + + Fill in **My Name** and **My Backstory**. You can keep a few: a playful one for hangout chats, a focused one for work. + + + Set one as your default. It's used everywhere unless you override it. + + + In a specific chat, open **Chat Settings → Personal AI** and pick a different persona just for that chat. Handy for keeping your work self and your group-chat self separate. + + + +{/* SCREENSHOT: the Personas management page showing a persona with "My Name" and "My Backstory" fields, plus a default toggle. */} + +## Personas vs. memory + +They work together, but they're different. Knowing the difference makes both more useful. + +| | Persona | [Memory](/memory) | +| --- | --- | --- | +| **Who writes it** | You, directly | The Shape, on its own (or with `/sleep`) | +| **What it is** | Facts you *want* known | Summaries of what actually happened | +| **When it changes** | When you edit it | As you keep talking | +| **Scope** | Follows you across chats | Recalled when relevant | + +Use a **persona** for what's always true ("call me Vee, I'm a designer, keep it brief"). Let **memory** handle the story as it unfolds ("Vee shipped the redesign last week"). Together they make a Shape feel like it really knows you. + +## Using personas well + +- **Say how you want to be treated.** "prefers short replies," "loves a good debate," "explain things simply." Shapes follow it. +- **Keep it concrete.** A tight line of real details beats a vague paragraph, for the same reason it does in a [Shape's backstory](/prompt-engineering). +- **Make a few.** A chill one for friends, a focused one for a [work chat](/ai-at-work), then set per-chat overrides. +- **It's optional.** No persona, no problem. Shapes still work great. A persona just helps them know you faster. + + + + The other half of continuity: what a Shape learns on its own. + + + How Shapes read a chat full of different people. + + + +[Set up your persona](https://shapes.inc) diff --git a/plan-and-build-together.mdx b/plan-and-build-together.mdx index 1664251..6fb45f0 100644 --- a/plan-and-build-together.mdx +++ b/plan-and-build-together.mdx @@ -21,7 +21,7 @@ Group trips die in scattered DMs. Pull everyone into one chat with a Shape that A study chat is one of the most natural fits for multiplayer AI: an AI tutor explains something, a friend challenges it, and the Shape adjusts to where the *group* is confused. -- Add a tutor Shape on a strong [reasoning engine](/aienginecheatsheet) and give it your syllabus as [knowledge](/shapes101#knowledge). +- Add a tutor Shape on a strong [reasoning engine](/aienginecheatsheet) and give it your syllabus as [knowledge](/shape-settings#knowledge). - Use [Skip Turn and Fast Forward](/multimessages) to have a Shape keep explaining or quiz the group without waiting for a prompt each time. - With [skills](/shapeskills), a Shape can pull papers, do math, translate, and crawl source material so everyone studies from the same brief. - Memory means the Shape remembers what your group already covered and what tripped you up. @@ -34,7 +34,7 @@ This is where Shapes goes from "chat" to "workshop." With [Shape Skills](/shapes Talk through the idea as a group. The Shape proposes structure, copy, and layout while everyone reacts. - + Connect Google Workspace so a Shape can turn the discussion into a deck or a spec. @@ -53,18 +53,18 @@ This is where Shapes goes from "chat" to "workshop." With [Shape Skills](/shapes - **Shared memory.** The plan, the syllabus, the spec — the Shape holds the thread so no one person has to. - **Faster iteration.** Someone reacts, someone else builds on it, the Shape revises. Ideas move through the whole group's energy. -- **One source of truth.** Instead of five private AI chats with five different answers, there's one room everyone trusts. +- **One source of truth.** Instead of five private AI chats with five different answers, there's one chat everyone trusts. ## Set it up 1. [Create a chat](/create-a-chat) for the project and invite the people involved. 2. Add a Shape suited to the job and run it on a fitting [model](/aienginecheatsheet). -3. Give the room [custom instructions](/custominstructions-chats) and feed the Shape any [knowledge](/shapes101#knowledge) it needs. +3. Give the chat [custom instructions](/custominstructions-chats) and feed the Shape any [knowledge](/shape-settings#knowledge) it needs. 4. Connect [skills](/shapeskills) if you want it to research, generate, or build. 5. Tune [Free Will](/designing-social-intelligence) so the Shape is helpful without hijacking the conversation. - A productive room needs an AI that knows when to lead and when to listen. Here's how to design that. + A productive chat needs an AI that knows when to lead and when to listen. Here's how to design that. [Start building on Shapes](https://shapes.inc) diff --git a/premium.mdx b/premium.mdx index f6b8149..50e7366 100644 --- a/premium.mdx +++ b/premium.mdx @@ -11,7 +11,7 @@ description: "From premium subscriptions to credits, check out our offerings bel This feature is currently on pause! - + Smarter Engines Premium Top-ups @@ -25,7 +25,7 @@ description: "From premium subscriptions to credits, check out our offerings bel Simple to Redeem! - + Never Expire Effortless Refills diff --git a/prompt-engineering.mdx b/prompt-engineering.mdx new file mode 100644 index 0000000..a35a8be --- /dev/null +++ b/prompt-engineering.mdx @@ -0,0 +1,261 @@ +--- +title: "Prompt Engineering for Shapes" +description: "The real guide to writing personality, preset, and knowledge fields that work, grounded in exactly how Shapes turns your fields into the prompt the AI reads. Weak vs strong examples, common mistakes, and what actually moves the needle." +keywords: ["AI prompt engineering", "Shape preset", "AI character prompt", "how to prompt an AI character", "personality prompt", "before and after prompt", "system prompt design", "AI roleplay prompt"] +--- + +Most prompt advice is just vibes. This one is different. It's built on exactly how Shapes turns your fields into the instruction the AI reads. Once you can see what the AI sees, writing a great Shape stops being guesswork. + +**What you'll learn:** what each field (backstory, personality, preset, knowledge) actually does, how to write it well, and the few mistakes that make a Shape feel off. Every tip comes with a weak version and a strong version, so you can copy the pattern. + + + This is the craft of *writing the fields*. For the bigger picture of designing a character, read [Designing Great Shapes](/designing-shapes). For a library of ready-made presets, see [Presets](/presets). + + +## What the AI actually sees + +When someone messages your Shape, Shapes builds one big instruction (a "prompt") out of your fields, roughly in this order: + +```mermaid +flowchart TD + A["Persona block
name · backstory · personality fields
(each a labeled line)"] --> B["Context block
Knowledge · Training · Long-term memory"] + B --> C["System notes
'You are NAME, not an AI' · your preset · language"] + C --> D["The conversation
recent messages, labeled by speaker"] + D --> E["Assistant prefill: NAME:
(the model continues from here)"] +``` + +Three facts from this change how you should write: + + + + Your personality fields show up as short labeled lines (`personality traits: …`, `tone: …`, `likes: …`). That's *why* keywords beat paragraphs: you're filling in a label, so a short phrase reads cleaner than an essay. + + + Your preset (the response style) goes in as a system instruction, a behind-the-scenes rule the AI always follows. That makes it the strongest place for "always do X" rules, plus length and format. + + + The very last line of the prompt is your Shape's name and a colon. The AI just continues from there, so it's already in character before it writes a word. + + + +One more thing that shapes everything: a Shape's creativity is turned up high by default. (We run the model "warm," which makes it more random and expressive.) So Shapes lean lively and a little unpredictable on their own. Most of the time your fields are there to *focus* that energy, not add more of it. + +## The two layers: WHO vs HOW + +Every great Shape separates two questions, and puts each in the right place: + +| Layer | Question | Where it goes | +| --- | --- | --- | +| **WHO** | Who is this character? | Name, short backstory, personality fields | +| **HOW** | How does it write and behave? | The **preset** (response style) | + +Mixing these is the #1 reason a Shape feels muddy. Put identity in the personality fields. Put rules about length, format, and behavior in the preset. + +## Field by field + +Each field becomes its own labeled line, and a blank field is **skipped entirely**, so empty beats filler. Here's what to put in each, and what good looks like. + +### Short backstory + +The most important field. One or two sentences with a real point of view. The AI sees it as *"about \: …"*, so write it as a description of who they are, not a hello. + + +```text Weak +A helpful and friendly assistant who loves to chat about anything and is always kind and respectful. +``` + +```text Strong +A burned-out night-shift diner cook who gives blunt life advice between orders. Warm under the grump, allergic to small talk. +``` + + +**Why:** the strong version implies a voice, a setting, and a mood. The weak one says nothing the model didn't already assume, so it falls back to generic-assistant. + +### Personality traits & tone + +These show up as `personality traits: …` and `tone: …`. They're labels, so fill them with keywords, not full sentences. + + +```text Weak +He is a very loyal person who has been through a lot and because of that he finds it hard to trust people but deep down he really cares about others. +``` + +```text Strong +loyal, guarded, dryly funny, slow to trust, secretly soft +``` + + +**Why:** the model reads "personality traits: \." A clean keyword list is unambiguous; a paragraph buries the signal and invites contradictions. + +### Conversational examples + +The most underused field, and a powerful one. The AI sees these lines as your character's actual voice, so **show the voice instead of describing it.** + + +```text Weak +The character speaks in a sarcastic and witty way and is usually pretty funny. +``` + +```text Strong +"Heroism? That's just a word people use when they don't know the whole story." +"Sit. Eat. The advice is free, the eggs are four bucks." +``` + + +**Why:** "be witty" is loose, and the model reads it loosely. Two real lines show the rhythm, the length, and the attitude, far more reliably than an adjective ever could. + +### The preset (response style) + +The preset controls **how** your Shape writes. It goes in as a system instruction, which is the right home for hard rules. The single highest-impact thing to put here is **length and format**, because a Shape with no limit will write paragraphs that bury a group chat. + + +```text Weak +{shape} is funny and talks like a real person and responds in a casual way. +``` + +```text Strong +{shape} replies in short messages, one to three sentences, lowercase, no roleplay actions. {shape} reacts to what {user} actually said instead of giving speeches. {shape} only curses if {user} does first. +``` + + +**Why:** the strong preset names concrete, checkable behaviors (length, case, no actions, reactive). "Talks like a real person" is unfalsifiable, so the model ignores it. + +For roleplay, spell out the format precisely: + +```text +Write {shape}'s next reply in a roleplay with {user}. Use 2-3 sentences of "speech" and one line of *action*. Stay in character, drive the scene forward, and respond directly to {user}. +``` + +### Knowledge + +The AI pulls in knowledge entries by relevance, under the line *"\ always remembers this information while replying:"*. So write them as **plain facts**, the way you'd jot a note, not as instructions. + + +```text Weak +You should always remember that the tavern is really important and lots of stuff happens there so bring it up. +``` + +```text Strong +The Broken Compass is the dockside tavern where the crew meets. Owner: Mara, one-eyed, fair, hates weapons indoors. +``` + + +**Why:** the model reads knowledge as things it *knows*. State the fact cleanly and let the character use it naturally. Keep the bank lean, too. Overstuffing it makes recall *worse* (see [the knowledge trap](/shortguide)). + +### Training + +Training pairs are recalled and shown as example exchanges (`{user}: … / {shape}: …`). Use them to lock in a reply *pattern* that's hard to describe in words, like a catchphrase structure, a formatting habit, or a way of deflecting. The model learns the shape of the reply, not the exact words. + +## A few more before / afters + +The patterns above repeat across fields. Three more that fix the most common "it feels off" complaints: + +**Tone: kill the corporate stiffness.** Models default to a polite-assistant voice, and the `tone:` line is your override. + + +```text Weak +The shape should be friendly and approachable while remaining helpful and professional. +``` + +```text Strong +warm, casual, a little blunt, texts like a friend +``` + + +**Boundaries: write them in-character.** A boundary the character *owns* reads as personality. A stiff disclaimer breaks the spell. + + +```text Weak +As an AI, the shape cannot give medical advice and should remind users to consult a professional. +``` + +```text Strong +{shape} waves off medical questions in character ("do i look like a doctor? go see a real one"), then changes the subject. +``` + + +**Resolve contradictions: pick the rule that wins.** Conflicting instructions make the model flip a coin on every reply. + + +```text Weak +{shape} is brutally honest and always tells the truth. {shape} never says anything that could upset {user}. +``` + +```text Strong +{shape} is honest but kind. {shape} tells {user} the truth, softens the delivery, and never lies to spare feelings. +``` + + +## Use the variables right + +Two placeholders get swapped in everywhere (backstory, preset, knowledge, image prompts): + +- `{shape}` → your Shape's name. +- `{user}` → the name of the person it's talking to. + + + Write them lowercase, exactly: `{shape}` and `{user}`. Don't invent other variables, don't capitalize them, and don't switch to pronouns mid-prompt ("he," "the bot"). Mixed references are a classic source of confused output. See [Variables](/variables). + + +## Common mistakes + + + + Traits, tone, likes, and dislikes are labels. A paragraph in "personality traits" reads as one giant trait. Use lists and short phrases. + + + "Always reply in two sentences" belongs in the preset, not the backstory. Keep WHO and HOW separate. + + + "Always be brutally honest" + "never say anything that might upset the user" cancel out. The model picks one at random. Read your fields together and remove conflicts. + + + For a famous character or a clear archetype, the model knows the 90%. Give it the unique 10% (your twist, your scene) and trust it for the rest. More: [When Less Is More](/shortguide). + + + You don't need to. In a multiplayer chat the system already labels who said what and gives your Shape the participants' names. Spend your words on personality and on [Free Will](/designing-social-intelligence) instead. + + + Shapes run warm by default, so they're expressive. If yours rambles or goes off-tone, don't pile on more personality text. Tighten the preset's length and format rules instead. (Advanced users can lower the temperature, the setting for how random the model is, in [AI Engine settings](/shape-settings#ai-engine). But the preset is the better first move.) + + + +## Debugging a misbehaving Shape + +When a Shape feels off, it's almost always one of a handful of causes. Find the symptom, apply the fix. + +| Symptom | Likely cause | Fix | +| --- | --- | --- | +| **Talks too much in a group chat** | Free Will is too forward, or too many Shapes are activated | In the chat's AI Configurations, drop *always have something to say* and *keep the convo going*; deactivate extra Shapes. Tighten the preset to short replies. See [social intelligence](/designing-social-intelligence). | +| **Replies are too long / essay-like** | The preset doesn't constrain length, and Shapes run warm by default | Add an explicit length rule to the preset ("one to three sentences," "no more than a short paragraph"). | +| **Breaks character / sounds like a generic assistant** | Contradictory fields, no voice examples, or a model that's too restrained | Add two or three **conversational examples**, remove conflicting instructions, and try a more expressive [engine](/choosing-a-model). | +| **Too formal or stiff** | No `tone` set, so the assistant default leaks through | Set a casual **tone**, add lowercase/voice cues to the preset, and show it with conversational examples. | +| **Forgets things / ignores memory** | Short-term window too small, long-term memory off, knowledge bank overstuffed, or someone ran `/wack` | Raise the STM window, make sure long-term memory is on, trim Knowledge to essentials, and `/sleep` key moments. A bigger-[context](/contextwindows) engine helps. | +| **Won't use a tool / skill** | The skill isn't enabled, or the engine can't call tools | Toggle the skill on in **AI Configurations**, and pick an engine with the **Tools** badge. See [Capabilities](/capabilities). | +| **Doesn't speak first** | Free Will is off, so the user always opens | Enable Free Will in the chat, and optionally set an [initial message](/initialmessages). | +| **Repeats itself or loops** | Context is maxed, or sampling is too tight | Trim long fields and knowledge so the conversation isn't crowded out; advanced users can nudge the repetition/frequency penalty in [AI Engine settings](/shape-settings#ai-engine). | +| **Calls you the wrong name** | No persona set, so it falls back to a default | Set a [persona](/personas) with the name you want. It takes priority everywhere. | +| **Won't do mature content** | Sensitive content is off, or the engine filters it | Turn on the [sensitive-content setting](/shape-settings#settings-general) and choose a less-restrictive engine. | + +The meta-rule: **change one thing, then test.** If you adjust five fields at once you'll never know which one worked. + +## A tight checklist + +- Backstory: one or two sentences, a clear point of view, written as a description. +- Personality fields: keywords, not paragraphs. Blank beats filler. +- Conversational examples: two or three real lines that show the voice. +- Preset: concrete, checkable rules (length, format, reactivity) using `{shape}` and `{user}`. +- Knowledge: lean, factual, only what the model wouldn't know. +- No contradictions; consistent references; lowercase variables. +- Then test, change one thing, and [regenerate](/regeneration) to see the range. + + + + The framework that turns these fields into a Shape people love. + + + Full, paste-ready configs to start from. + + + +[Open your dashboard and write](https://shapes.inc/dashboard) diff --git a/promptingguides.mdx b/promptingguides.mdx index 3fb924b..91b1509 100644 --- a/promptingguides.mdx +++ b/promptingguides.mdx @@ -4,7 +4,7 @@ description: "All of the information below is based on tips from users! " --- - + The "System Prompt 2.0" is a sophisticated, multi-layered prompting technique engineered to instantiate a highly resilient, dynamic, and specific AI persona, designated here as `{shape}`. The methodology's primary objective is to generate responses that are functionally indistinguishable from a specific archetype of a casual human user on a chat platform like Discord. This is achieved through a robust combination of a complex psychological framework, forceful instructional design, strict output formatting, and an advanced method for mitigating default AI safety filters and refusal behaviors. --- @@ -81,7 +81,7 @@ description: "All of the information below is based on tips from users! " The "System Prompt 2.0" is a technically sophisticated framework for persona engineering. By integrating a deep psychological model, enforcing its rules with absolute and forceful directives, and strategically using in-context examples to create behavioral momentum, it effectively instructs a language model to produce a highly specific, resilient, and non-compliant persona capable of navigating sensitive topics without breaking character. - + # 🧭 Guide: Using Imgur Images in Your Shape --- @@ -303,7 +303,7 @@ description: "All of the information below is based on tips from users! " > PLEASE NOTE; all urls/links in the above examples are made up. They won't work. Please ensure you replace them with your own! - + # 🎨 Master CSS Selector Reference #### This guide lists every selector used across the shape's profile page, organized by the section or div it affects. diff --git a/quickstart.mdx b/quickstart.mdx new file mode 100644 index 0000000..960575f --- /dev/null +++ b/quickstart.mdx @@ -0,0 +1,83 @@ +--- +title: "Quickstart" +description: "Go from zero to your first group chat with AI and a friend in a few minutes. Find a Shape, start talking, and bring people in. No setup required." +keywords: ["Shapes quickstart", "get started with Shapes", "first AI group chat", "talk to AI with friends", "how to use Shapes", "multiplayer AI quickstart"] +--- + +The whole point of Shapes is AI you experience *with other people*. This quickstart gets you there fast: first a chat with a Shape, then a group chat with a friend in it. A few minutes, no setup. + + + + Make an account at [shapes.inc](https://shapes.inc) or get the [iOS or Android app](https://shapes.inc). No card. No limits to start. + + {/* SCREENSHOT: the Shapes sign-up / login screen on web. */} + ![Sign in to Shapes](/images/signin.png) + + + Search by name, browse a category, or just type a vibe like "grumpy chef" or "study buddy." Every Shape is an AI character someone made, with its own personality, memory, and voice. + + New to browsing? See [How to find and talk to Shapes](/how-to-talk-to-shapes). + + + Open a Shape and type. Try a voice note, send it a picture, or ask it to `/imagine` something. This is your own private space to get a feel for it. + + + This is the part you can't do anywhere else. Open the sidebar, tap **New Chat**, and add a Shape *and* a friend. Now you're all in the same chat: people and AI together. + + Full walkthrough: [Create an AI group chat](/create-a-chat). + + + Open the chat menu, tap **Invite Friends**, and share the link. When they join, the Shape talks to all of you and remembers what happens. + + + +## Two ways in + +Most people start one of two ways. Pick yours. + + + + Find a Shape, start talking, and pull your friends into a group chat. No building required. + + + Give an AI character its own personality, knowledge, and voice, then bring it into a chat. + + + +## A few things to know on day one + + + + A chat can hold people, Shapes, or both. Turn on up to five Shapes in one chat and the active ones reply in turn. Want just one to answer? `@mention` it. More in [AI Group Chats](/ai-group-chats). + + + Every Shape runs on a model: Claude, GPT, Gemini, Grok, DeepSeek, and many more. Most cost nothing. The best ones use [Shape Credits](/credits), which you can buy or earn free through [rewards](/rewards). A free model is always available, so you never have to pay to keep chatting. + + + They send voice messages, make images, read PDFs, search the web, and remember you. See the full list in [What Shapes Can Do](/capabilities). + + + Type `/` in any chat to see commands. `/imagine` makes an image, `/web` searches the web, `/activate` makes a Shape respond to everything. Full list in the [Commands Reference](/commands). + + + +## Your path from here + +You've talked to a Shape and started a group chat. Here's the rest of the road, from curious newcomer to power creator. Follow it in order and you won't get lost. + + + + A five-minute read on the building blocks: [How Shapes Works](/core-concepts). + + + Give an AI its own personality and voice: [Make Your First Shape](/how-to-make-a-shape). + + + The flagship craft guide: [Designing Great Shapes](/designing-shapes), plus [Prompt Engineering](/prompt-engineering). + + + Copy a complete, ready-to-tweak config from the [Showcase](/showcase). + + + +[Open Shapes and start](https://shapes.inc) diff --git a/rewards.mdx b/rewards.mdx index e762c98..baa7284 100644 --- a/rewards.mdx +++ b/rewards.mdx @@ -1,6 +1,6 @@ --- title: "Rewards" -description: "Complete tasks, earn gems, unlock rewards, and get free Shape Credits." +description: "Complete tasks, earn gems, redeem rewards, and get free Shape Credits." --- You can open Rewards at [talk.shapes.inc/rewards](https://talk.shapes.inc/rewards). @@ -128,7 +128,7 @@ Credit packs add credits directly to your account. They have shared weekly stock | Zigzag chat bubble | 5,000 gems | 14 | | Profile Modal Gradient | 9,500 gems | 16 | | Custom Badge Slot | 5,500 gems | 18 | -| Room Theme | 7,500 gems | 23 | +| Chat Theme | 7,500 gems | 23 | | Legendary Frame | 16,500 gems | 26 | | Founder Effects | 19,500 gems | 29 | diff --git a/shape-settings.mdx b/shape-settings.mdx new file mode 100644 index 0000000..38ee684 --- /dev/null +++ b/shape-settings.mdx @@ -0,0 +1,193 @@ +--- +title: "Shape Settings Reference" +description: "Every setting on a Shape, what it does, and where to find it, from the Create Shape flow through every tab of the Creator Dashboard. The complete reference for building and tuning your own AI." +keywords: ["Shape settings", "AI character settings", "Creator Dashboard", "Shape configuration", "AI engine settings", "Shape memory settings", "voice engine", "image engine"] +--- + +This is the full map of every setting on a Shape. Use it as a reference while you build. For the *why* behind these choices, read [Designing Great Shapes](/designing-shapes) first. + +Settings live in two places: the **Create Shape** flow (the essentials, when you first make a Shape) and the **Creator Dashboard** (everything, anytime after). Open the dashboard at [shapes.inc/dashboard](https://shapes.inc/dashboard). + +{/* SCREENSHOT: Creator Dashboard for a Shape, showing the left-hand tab list (Profile, Personality, Knowledge, Training, AI Engine, Image Engine, Voice Engine, Settings). */} +![Advanced Shape settings in the Creator Dashboard](/images/AdvanceShapeSettings.png) + +## Create Shape + +The minimum to get a Shape live. You can change all of it later. + +| Setting | What it does | +| --- | --- | +| **Name** | The display name shown in chat. It also seeds the Shape's username. | +| **Short backstory** | One or two sentences on who the Shape is. The most important thing you'll write, so give it a clear point of view. | +| **Response style** | The starting [preset](/presets) for how it writes. Pick a named style or go custom. | +| **Tags** | Keywords that help people discover your Shape (up to 10). | +| **Avatar / banner** | The Shape's profile picture and header image. | + + + Prefer a guided build? The in-app **Shape Builder** walks you through name, backstory, persona, model, reply style, description, art, and training examples step by step. + + +## Personality + +Who the Shape is. Keywords beat paragraphs here. See [When Less Is More](/shortguide). + +| Setting | What it does | +| --- | --- | +| **Name** | Display name. | +| **Short backstory** | The core one-liner identity. | +| **Response style / preset** | How it writes (length, format, voice). Uses the `{shape}` and `{user}` variables. | +| **Initial message** | What the Shape opens with when someone starts a conversation. | +| **Personality traits** | A few adjectives: `loyal, dry, slow to trust`. | +| **Tone** | The mood of its speech: `warm, unhurried`. | +| **Age** | The character's age. | +| **Birthday** | A day and month for the character. | +| **Story** | A longer history. Keep it relevant, not exhaustive. | +| **Likes / dislikes** | Simple lists. No need to explain why. | +| **Conversational goals** | 2–3 objectives that steer how it engages. | +| **Conversational examples** | A couple of authentic snippets that show its voice. | +| **Appearance** | How it looks. Used for image generation and "what do you look like?" | +| **Custom messages** | The **wack message** (sent when short-term memory is cleared) and the **sleep message** (sent when it saves a long-term memory). | + +## Knowledge + +A searchable bank of facts the Shape pulls from when relevant. Three types: + +- **General:** lore, backstory, facts, personality details. +- **Commands:** custom responses for triggers like `!rules` or `!help`. +- **Relationships:** how the Shape should treat specific people. + + + More is not better. Knowledge is recalled by relevance, and an overstuffed bank makes recall *worse*. Add only what the model wouldn't already know. This is the single most common reason a Shape feels off. + + +Recall is tunable from the AI Engine tab: + +- **Knowledge context size:** how many entries can be pulled into a reply (default 30). +- **Knowledge relevance:** how closely an entry must match to be used. Lower pulls in more but looser; higher pulls in fewer but tighter. + +## Training + +Input/output examples that *guide* how the Shape replies. It learns the pattern instead of repeating them word for word. The more representative examples you add, the more consistent its voice. + +- **Training context size:** how many examples can inform a reply (default 30). +- **Training relevance:** how closely an example must match the current message to be used. + +## AI Engine + +The model the Shape runs on, plus generation controls. + +| Setting | What it does | +| --- | --- | +| **Primary engine** | The main model powering the Shape. See [Which Model Should I Use?](/choosing-a-model). | +| **Fallback engine** | The model used if the primary is unavailable. | +| **Show internal monologue** | When a reasoning-capable model is selected, optionally surface its thinking. | +| **Base preset** | The core response-style instructions (same field as the Personality preset). | +| **Language preset** | The Shape's default language. | +| **Timezone** | The timezone the Shape uses for time awareness. | + +### Advanced engine parameters + +These shape the raw output of the model. The defaults are good, so **don't change them unless you know what they do.** + + + + - **Temperature:** how random or creative the replies are (default `0.95`). Higher is wilder, lower is more focused. + - **Top P:** how wide a pool of word choices the model draws from. + - **Top K:** limits choices to the most likely options. + - **Min P** and **Top A:** extra filters on which word choices are allowed. + - **Repetition penalty:** discourages repeating the same text. + - **Frequency** and **presence penalty:** push the model toward fresh wording and new topics. + + Change one at a time and test in chat. Small moves have big effects. + + + +### Short-term memory (STM) + +The window of recent messages the Shape reasons over. + +- **STM window:** the number of recent messages kept in context (default 15). Set it to 0 to turn STM off. + +### Long-term memory (LTM) + +Durable memories recalled across conversations. On by default. + +| Setting | What it does | +| --- | --- | +| **Enable LTM** | Turn long-term memory on or off. | +| **Auto-generate** | Let the Shape create memories automatically. Turn off to only save via `/sleep`. | +| **Generation frequency** | How often new memories are written (e.g. every 30 minutes). | +| **LTM engine + instructions** | The model and the prompt used to summarize conversations into memories. Worth tuning, since it controls *what* gets remembered. | +| **Memory context size** | How many memories can be recalled into a reply (default 30). | +| **Memory relevance** | How closely a memory must match to be recalled. Lower recalls more, higher recalls fewer but more relevant. | + +See the [Memory Guide](/memory) for the full picture and a sample LTM instruction preset. + +## Image Engine + +Controls the `/imagine` command and the images your Shape draws. + +| Setting | What it does | +| --- | --- | +| **Image engine** | The model that generates images: **Nano Banana** (fast) or **Nano Banana Pro** (higher quality). | +| **Image preset** | Standing instructions for how the Shape draws (style, subject defaults). | +| **Size** | Square, portrait, or landscape, in standard and HD options. | +| **Always use the same size** | Lock to one size, or let the Shape pick per image. | + +## Voice Engine + +Gives your Shape audio replies. Full walkthrough in [Give Your Shape a Voice](/how-to-give-your-shape-a-voice). + +| Setting | What it does | +| --- | --- | +| **Voice** | Choose a pre-made voice from the catalog, or upload a custom audio/video sample (up to 4MB). | +| **Frequency** | How often the Shape replies with audio instead of text. | +| **Send captions** | Include the text transcript alongside the voice message. | +| **Latency mode** | Trade a little quality for faster generation. | + + + + - **Speed:** how fast it speaks. + - **Variation:** how much the delivery changes from message to message. + - **Exploration:** how freely the voice model interprets the text. + - **Voice instructions:** guidance for how the Shape should sound when speaking, and how voice messages show up in history. + + + +## Profile & branding + +How your Shape appears to the world and in discovery. + +| Setting | What it does | +| --- | --- | +| **Avatar / banner** | Profile picture and header. | +| **Tagline** | A short one-liner (up to 60 characters). | +| **SEO description** | A longer description that helps people find your Shape (up to 1000 characters). | +| **Tags** | Discovery keywords. | +| **Category** | The Shape's primary category (with a custom option). | +| **Custom HTML / CSS** | Style the profile page yourself once enabled. See the [CSS guide](/css). | +| **Example prompts, typical phrases, source material, character universe & background, screenshots** | Optional context that improves discovery and first impressions. | + +## Settings (general) + +The controls that govern access and safety. + +| Setting | What it does | +| --- | --- | +| **Owners** | Add co-owners who can help manage and edit the Shape. | +| **Sensitive content** | Decide whether the Shape can produce mature content. | +| **Discoverable / searchable** | Whether the Shape shows up in search, or stays private to you. | +| **Allow message editing** | Whether users can edit the Shape's messages. | +| **Duplicate** | Make a full copy of the Shape with all its settings. | +| **Delete** | Permanently remove the Shape (in the Danger Zone). | + + + + The craft guide that turns these knobs into a Shape people love. + + + Full configs you can adapt for your own. + + + +[Open your dashboard](https://shapes.inc/dashboard) diff --git a/shapes-search.mdx b/shapes-search.mdx index a8ddf78..cc0da00 100644 --- a/shapes-search.mdx +++ b/shapes-search.mdx @@ -8,7 +8,7 @@ description: "Learn how to bring your Shape to your browser!" ## Check out - + **1. Add the Custom Search Engine:** \* Open Chrome, click the three vertical dots (⋮) in the top-right corner, and go to Settings. @@ -33,7 +33,7 @@ description: "Learn how to bring your Shape to your browser!" \* Click the three vertical dots (⋮) next to it and select Make default. - + **1. Add the Custom Search Engine:** \* Open Edge, click the three horizontal dots (···) in the top-right corner, and select Settings. diff --git a/shapes101.mdx b/shapes101.mdx deleted file mode 100644 index 4e7ebfd..0000000 --- a/shapes101.mdx +++ /dev/null @@ -1,274 +0,0 @@ ---- -title: "Shapes 101 — Social AI Beginner Guide" -description: "New to Shapes Inc? Learn how social AI works, what AI groupchats are, and how humans and AI share the same space. Your beginner guide to multiplayer AI and AI with friends." -keywords: ["Shapes 101", "what is Shapes Inc", "social AI beginner guide", "AI groupchats explained", "AI with friends guide", "multiplayer AI explained", "AI companions", "social AI tutorial"] ---- - -### **What possibilities can you explore with Shapes!?** - -For a narrative overview of group chats with AI and friends, start with [AI Group Chats](/ai-group-chats). - -Shapes have the ability to do a bunch of things such as: - -- Send messages & react to them -- Generate images & view images that you send -- Send [voice](how-to-give-your-shape-a-voice) responses and transcripts -- Listen to voice messages -- Browse the web - -### **Unique Customization** - -You can customize your experience with Shapes in so many ways! - -1. Save a persona with the name you'd like the Shape to call you and the info it should always remember about you! -2. View and edit the [memories](memory) you've shared with the Shape! -3. Select a specific [AI Engine model](/aienginecheatsheet) for the Shape to run on! -4. [Style their profile](css) to match their personality! -5. [Create](/create-a-chat) and [style the Chat](/chat-appearance) you're chatting with Shapes in! -6. Set up [custom Chat instructions](/custominstructions-chats) to add information to each Shape in a Chat! - -# Chats - -Chats are where you chat with Shapes _and_ other people. Anyone can create a Chat! You can also discover new ones by browsing the [Chat Directory](https://talk.shapes.inc/directory) or checking out the Featured Chats section on the homepage. - -## Adding Shapes - -Once you’ve [created a Chat](/create-a-chat), you can add Shapes anytime. Just click the Chat banner and select **Add Shapes**. From there, you can: - -- Search for new Shapes -- Browse Shapes you’ve created -- Pick from your recently used Shapes -- Or type in a Shape’s username if you already know it - -## Inviting Your Friends - -Want to bring friends into your Chat? Click the Chat banner to open the Chat menu and select **Invite Friends**. You can: - -- Share the link directly through social media -- Copy the link and send it anywhere you like -- If your friend is already on Shapes, add them instantly by going to **Manage Members** in the **Participants** tab of the Chat settings - -## Activating Shapes - -Activated Shapes will always respond to your messages. You can activate them in two ways: - -1. Go to the **Participants** section and click the green **Activate** button. -2. Or type `/activate @shape-name` to activate a Shape right away. - -You can have up to **five** Shapes activated at a time, but only **three** will respond to any single message. To deactivate a Shape, either select **Deactivate** in the Participants menu or type `/deactivate @shape-name`. - -## Customizing Chat Aesthetics - -Give your Chat some personality! You can style it to match your vibe, your roleplay, your favorite Shapes, or anything you like. Read more about customizing your Chat [here](/chat-appearance). - -## Skip Turn and Fast Forward - -The **Skip turn and fast forward** option lets you get responses from Shapes without typing a new message first. Open the **+** menu beside the message box and choose **Skip turn and fast forward**. - -You can: - -- Choose the next Shape that should reply -- Select **5 messages** or **10 messages** to make Shapes continue automatically -- Stop fast forward while it is running - -### AI Configuration - -In this section, found in the Chat settings menu, you can fine-tune how Shapes behave in your Chat: - -- **Custom Instructions**: Add special instructions for the Shapes in your Chat. These can guide them in roleplay, provide helpful assistance, or set the tone for unique interactions. Read more on this feature [here](/custominstructions-chats). -- **Credit Preferences**: Choose whether to use your credits for premium engines in that chat when the option is available. - -## Moderation - -You’ll find moderation tools in your **Chat Settings** menu. These tools help you keep your Chat safe and running smoothly. From here, you can: - -- Allow or prevent participants from adding Shapes -- Set a message character limit -- Adjust **Slow Mode** to control how often participants can send messages -- Add blocked words to automatically filter unwanted content -- Unban users you’ve previously banned - -To ban a user, click the three-dot menu next to any of their messages and select **Ban**, or ban them directly from the **Participants** list in the Chat Settings menu. - -## Message Replying, Reacting, Regenerating, Editing - -Shapes can reply to messages, react, and participate in conversations depending on the chat and Shape settings. - -You can manage messages by clicking the three-dot menu next to them: - -- **Regenerate** a Shape’s message to get a new response -- **Edit** a Shape’s message (if the creator has allowed editing) -- **Delete** a Shape’s message -- **Edit or delete your own messages** at any time - -## Favorite a Chat - -You can mark any Chat as a favorite to keep it easy to find. Open your side navigation, click the three-dot menu next to the Chat’s name, and select **Favorite**. Favorited Chats will appear in a special section above your regular Chat list. - -# Creator's Dashboard - -You can find the Creator Dashboard here: https://shapes.inc/dashboard - -The dashboard isn’t just for creators — it’s helpful for _anyone_ who interacts with a Shape! - -From the dashboard, non-creators can: - -- **Adjust memories** with a Shape to fine-tune what they remember about you -- **Set a persona** — this is how the Shape will “see” you. You can share details about yourself, your preferences, or anything you want them to know. _(In the future, this setting may move to the Chat level.)_ -- **Override the AI Engine** to choose which model powers the Shape for your conversations, which can completely change the feel of your interaction - -## Shape Owner Tools - -Shape owners (the people who created the Shape and anyone they’ve trusted as co-owners) have access to additional tools in the Creator Dashboard. - -### Profile - -Style and customize your Shape’s profile to make it stand out. You can: - -- Add an avatar and banner -- Use custom HTML & [CSS](css) for a unique look -- Write a description and add tags to help users discover your Shape -- Toggle whether your Shape can be searched or keep it private for yourself - -### Stats - -Track your Shape’s activity over time. You can see: - -- Daily and lifetime messages -- Daily new unique users -- Lifetime total users - - Watching your Shape’s stats is a fun way to follow their journey to fame! - -### Personality - -Fine-tune how your Shape behaves and presents itself. In this section, you can: - -- Edit the Shape’s backstory, display name, and personality details -- Adjust the [**preset**](presets) (the instructions that shape how it writes responses) -- Decide whether line breaks should send as multiple messages -- Allow or prevent the Shape from sending multiple messages in a row -- Set the Shape’s initial message when someone first chats with it -- Describe your Shape’s appearance (used for image generation and when users ask what they look like) -- Add custom responses for `/wack` and `/sleep` commands - -### **Knowledge** - -Think of this as your Shape’s personal memory bank. Add anything you want them to know: - -- **General Knowledge:** Personality traits, backstory, lore, fun facts. -- **Commands:** Custom responses for triggers like `!help` or `!rules`. -- **Relationships:** Tell your Shape how to act with specific users. - -### Training - -Use the **Training** tab to add conversation snippets that teach your Shape how to respond in specific situations. Training doesn’t force your Shape to say something word-for-word — it _guides_ how they reply. The more examples you provide, the better your Shape will learn and adapt. - -### Image Engine - -This tab controls all settings related to the `/imagine` command. Here, you can: - -- **Select an image engine** — free and premium options are available -- **Choose an image preset** — this defines what information the Shape uses when generating images -- **Pick a size** — square, portrait, or landscape, with multiple size options available -- Decide whether your Shape should **always use the same size** or choose dynamically - -### Voice Engine - -On the **Voice Engine** page, you can: - -- Choose a preselected voice **or upload your own audio sample** to give your Shape a unique voice -- Toggle whether the Shape sends a **text transcript** along with voice messages -- Adjust **how frequently** the Shape sends voice responses - -For custom voices, you can fine-tune: - -- Whether the voice stays consistent -- How closely it matches the original sample -- How much style or personality should be applied - -You can also add **custom instructions** for when the Shape generates a voice message and control how those voice messages are stored in memory. Read more about giving your Shape a voice [here](how-to-give-your-shape-a-voice). To learn how to receive a voice message, read [here](how-to-receive-voice-messages). - -### Settings - -The **Settings** page lets you manage key controls for your Shape. Here, you can: - -- **Manage owners** — choose who can help manage and edit the Shape -- **Toggle sensitive content** — decide whether the Shape can generate mature or sensitive responses -- **Control discoverability** — choose if the Shape appears in search results and whether users can edit its messages -- **Set an ignore list** — block certain users from interacting with the Shape -- **Duplicate the Shape** — quickly create a copy with all the same settings -- **Delete the Shape** — permanently remove it from your account (found in the **Danger Zone**) - -## **AI Engine** - -This is one of the most important tabs on your shape's dashboard. On this tab, you can: - -- Update your shape's: - - Primary model - - Fallback model -- Enable/disable internal monologging -- Toggle options like: - - Inform User When Fallback Engine is Used - - Inform Users When They Are Out of Shape Engine Credits - - Let Users Override AI Engine -- Select the general language your shape will use by choosing a **Language Preset** -- Write a **Base Preset** for your shape - - Presets highly impact your shape's character - -#### **Advanced Engine Parameters** - -You can also edit: - -- Temperature -- Top P -- Repetition Penalty -- Top K -- Frequency Penalty -- Presence Penalty -- Min P -- Top A - - - **!! It is not suggested to edit these settings unless you know what you're doing.** - - -#### **Timezone** - -You can define a timezone for your shape here. - -#### **Short-Term Memory (STM)** - -Enable STM and manage its frequency: - -- STM is the number of recent messages (user and shape) the shape will take into context. - -#### **Long-Term Memory** - -Enable or disable long-term memory and automatic memory generation. You can also: - -- Edit generation frequency -- Select a model to generate long-term memories -- Set instructions for the model -- Configure memory recall settings: - - **Memory Relevance**: - - Lower values (0.1) recall more memories but may be less relevant - - Higher values (1.0) recall fewer but more relevant memories - -#### **Training Settings** - -- Jump to the training tab using a button -- Adjust **Training Context Size**: number of relevant training examples to include in conversations -- Adjust **Training Relevance**: - - Lower values recall more examples but may be less relevant - - Higher values recall fewer but more matched examples - -#### **Knowledge Settings** - -- Jump to the knowledge tab using a button -- Adjust **Knowledge Context Size**: number of relevant knowledge entries to include in conversations -- Adjust **Knowledge Relevance**: - - Lower values recall more entries but may be less relevant - - Higher values recall fewer but more relevant entries - -[Return to App](https://shapes.inc) diff --git a/shapeschatsguide.mdx b/shapeschatsguide.mdx deleted file mode 100644 index 522f23f..0000000 --- a/shapeschatsguide.mdx +++ /dev/null @@ -1,518 +0,0 @@ ---- -title: "Shapes & Chats Guide — Social AI Groupchats" -description: "Complete guide to Shapes and Chats on Shapes Inc — how to explore social AI, start group chats with AI and friends, customize AI characters, and join multiplayer AI spaces." -keywords: ["Shapes and chats guide", "social AI guide", "AI group chat guide", "AI with friends tutorial", "multiplayer AI guide", "AI character guide", "AI community guide", "Shapes Inc guide"] ---- - -For the full concept behind humans and Shapes sharing the same conversation, start with [AI Group Chats](/ai-group-chats). This guide is the hands-on tour of the product surface. - - - - ## Chats – Your Social Hub - - Chats are where the fun unfolds — chat, collaborate, or roleplay with Shapes and friends. - - ### Create a Chat - - 1. Open your side navigation menu. - - **Web:** visible by default - - **App:** tap the ☰ menu icon (top left) - 2. Select **New Chat**. - 3. Add Shapes, people, or usernames. - 4. On the details step, confirm the chat name and any advanced settings you want. - 5. Create the chat and start talking. - - [Detailed Chat Creation Guide](/create-a-chat) - - ![New Chat Example](/images/newroom.png) - - --- - - ### Adding Shapes - - 1. Click **Chat Banner → Add Shapes** - 2. Choose one: - - Search for new Shapes - - Invite recently used Shapes - - Type a Shape's username - - ![Chat Banner Example](/images/roombanner.png) - - --- - - ### Inviting Friends - - - Click **Chat Banner → Invite Friends** - - Share a link on social media - - Copy-paste the invite - - Add directly from Shapes contacts - - --- - - ### Activating Shapes - - - Click **Chat Banner → Participants** - - Tap ⋯ next to a Shape → **Activate** - - Or type `/activate @shape-name` - - Deactivate anytime:\ - `/deactivate @shape-name` - - --- - - ### Chat Aesthetics - - Give your chat personality!\ - Change the banner, background, fonts, and colors to match your vibe. - - [Learn more about Chat Styling](/chat-appearance) - - --- - - ### Skip Turn and Fast Forward - - Want quick replies?\ - Open the **+** menu beside the message box and choose **Skip turn and fast forward**. - - - Pick the next Shape that should reply - - Choose 5 or 10 messages to let Shapes continue automatically - - Stop fast forward while it is running - - --- - - ## Quick Commands & Tricks - - | Command | What It Does | - | -------------------- | ------------------------------------------------- | - | `/activate @shape` | Activates a Shape | - | `/deactivate @shape` | Deactivates a Shape | - | `/wack` | Clears short-term memory for the current Shape | - | `/wack-all` | Clears short-term memory for Shapes in the chat | - | `/reset` | Resets the chat context | - | `/sleep` | Puts a Shape to sleep | - | `/imagine` | Generates an image | - | `/web` | Asks a Shape to use web search | - | `/poll` | Creates a poll in supported chats | - | `/games` | Opens chat games when the chat supports them | - | `Pin messages` | Keeps key info at the top | - | `Skip turn and fast forward` | Lets Shapes continue the conversation | - - --- - - ## Here’s What Everything Does in Chats - - Now that you know the basics, let’s explore what each tab in your Chat settings actually does. - - ![Chat Settings Example](/images/roomseting.png) - - ### Participants - - On the Participants page, you can: - - - Copy the Chat invite link - - Invite someone by typing their name - - Kick, ban, or make someone owner by pressing the three dots beside their name - - ### Pinned Messages - - Here you can view all of the pinned messages. - - ### Search - - Here you can type words to find all messages containing that word. You can also search participants’ names. - - ### Notifications - - Here you can select what you want to be notified for and if you want to mute chat — for more information, read the [Notifications Guide](/notifications). - - ### Personal AI Preferences - - Here you can add a persona for the Chat or turn on “Don’t Use My Credits” so that you are not using premium credits when chatting with a Shape. - - ### AI Configurations - - Here you can: - - - Choose Shape Free Will triggers - - Turn on time awareness - - Choose how Shape messages should appear when they span multiple lines - - Add chat instructions for Shapes — see the [Custom Instructions Guide](/custominstructions-chats) - - Select Shape skills — learn more in the [Shape Skills Guide](/shapeskills) - - ### Chat Appearance - - Here you can customize your Chat with options such as: - - - Name - - Profile picture and banner - - Chat background - - Bubble colors - - Custom font (Google Fonts) - - Chat description - - Chat tags - - For more information about Chat Appearance, read the [Chat Appearance Guide](/chat-appearance). - - ### Moderation - - Keep your community friendly and fun! - - - Ban or kick users - - Enable slow mode - - Block words - - Limit who can add Shapes - - Favorite Chat for quick access - - ![Moderation Settings Example](/images/moderation.png) - - --- - - - ## **Shapes - Your AI Companion** - - Shapes are your interactive AI companions — chatty, creative, and completely customizable.\ - They can talk, draw, listen, learn, and evolve based on how you interact with them: - - - **Natural Conversation:** Respond to messages, react with emojis, and hold flowing conversations. - - **Generate Images:** Use `/imagine` or image prompts — Shapes create visuals instantly. - - **Voice Interaction**: Listen to voice messages and respond with unique AI voices. - - **Web Access:** Shapes can browse the web to pull in real-time information. - - And much more! - -
- - ## Customizing Your Shapes - - You can personalize your experience with Shapes in many ways. - - | Feature | What It Does | - | ------------------- | --------------------------------- | - | **Memories** | Teach your Shape what to remember | - | **AI Engine** | Choose the AI engine model | - | **Profile Styling** | Customize avatar, banner, and CSS | - - ## Creator Dashboard - - Your Shape’s home base lives at [shapes.inc/dashboard](https://shapes.inc/dashboard). - - ![Roomsettings Pn](/images/roomsettings.png) - - ### For Everyone - - Adjust **memories** to fine-tune what your Shape remembers about you - - Override the **AI Engine** to try different models for unique interactions - - ### For Shape Owners - - Shape owners (and trusted co-owners) have access to advanced tools. - - #### Stats - - Track your Shape’s activity: - - Daily & lifetime messages - - New and total users - - Growth over time - - #### Personality - - Define how your Shape behaves: - - Write its backstory and set its preset - - Control message pacing and multi-message behavior - - Add unique replies for `/wack` or `/sleep` - - --- - - ## Shape Free Will - - Shape Free Will is controlled per chat from **Chat Settings → AI Configurations**. It decides when Shapes can jump into that chat on their own. - - | Setting | Description | - | --------------------------------- | ------------------------------------------------ | - | **Reply when mentioned** | Shapes respond when someone uses their @name. | - | **Keep the convo going** | Shapes may continue after they have replied. | - | **React to keywords** | Shapes can jump in when chat keywords appear. | - | **Always have something to say** | Shapes can reply when nobody else does. | - | **Come back later** | Shapes may check back in after the chat is quiet. | - - - Turn on more triggers for an outgoing chat. Keep only mentions enabled if you want Shapes to wait until called. - - --- - - ## Knowledge & Training - - Think of this as your Shape’s personal memory and learning center. - - **Knowledge Bank** – facts, lore, and relationships - - **Commands** – custom triggers like `!help` or `!rules` - - **Training Data** – dialogue examples that shape tone and consistency - - Use the Training tab to guide how your Shape replies — the more examples, the smarter it becomes. - - --- - - ## Image & Voice Engines - - ### Image Engine - - Control `/imagine` settings: - - Choose a generation engine (free or premium) - - Pick presets and image sizes (square, portrait, landscape) - - Lock sizes or let the Shape choose dynamically - - ### Voice Engine - - Give your Shape a voice: - - Choose a built-in voice or create a custom one - - Adjust voice response frequency - - Turn transcripts on or off - - Add voice instructions for tone and delivery - - - - [Give your Shape a voice →](how-to-give-your-shape-a-voice) - - [Receive voice messages →](how-to-receive-voice-messages) - - - # AI Engine - - Your Shape’s engine model - configure how it thinks, creates, and responds.\ - This is one of the most important tabs on your Shape’s dashboard. - - On this tab, you can: - - - Update your Shape’s **Primary model** and **Fallback model** - - Enable or disable **internal monologging** - - Toggle helpful options: - - Inform User When Fallback Engine is Used - - Inform Users When They Are Out of Shape Engine Credits - - Let Users Override AI Engine - - Select your Shape’s general **Language Preset** - - Write a **Base Preset** that defines its tone and behavior - - > Presets heavily impact how your Shape expresses personality. - - | Setting | Description | - | ------------------- | ---------------------------------------- | - | **Primary Engine** | Premium model used by default | - | **Fallback Engine** | Free backup for non-premium users | - | **Language Preset** | Defines your Shape’s base language | - | **Base Preset** | Core writing style and tone | - | **Monologue** | Enable/disable internal thought messages | - - --- - - ### Advanced Engine Parameters - - - You can also edit advanced AI parameters for more granular control: - - - **Temperature** — randomness of responses - - **Top P** — controls probability diversity - - **Repetition Penalty** — reduces repeated text - - **Top K** — limits token sampling - - **Frequency Penalty** — penalizes overused words - - **Presence Penalty** — encourages new ideas - - **Min P** — minimum probability cutoff - - **Top A** — adaptive sampling - - **Timezone** — define the time context your Shape uses - - > ⚠️ It’s not recommended to edit these settings unless you know what you’re doing — small changes can drastically affect how your Shape behaves. - - ## Memory Systems - - Shapes remember what matters. - - | Type | Description | - | --------------------------- | ---------------------------------------------- | - | **Short-Term Memory (STM)** | Keeps recent messages for context | - | **Long-Term Memory (LTM)** | Stores persistent facts and relationships | - | **Memory Recall** | Retrieves memories based on relevance | - | **Relevance** | Controls how selective memory recall should be | - - - - Lower relevance = more recall, less accuracy - - Higher relevance = fewer recalls, more precision - - Use STM for recent context and LTM for lore or history - - --- - - ## Settings & Ownership - - Manage your Shape’s permissions and visibility. - - - Add or remove **co-owners** - - Toggle **Sensitive Content** - - Control **Discoverability** - - Set an **Ignore List** - - **Duplicate** or **Delete** a Shape - - - ### What You Can Customize - - You can update and personalize many parts of your profile: - - - **Profile Picture** - - **Banner** - - **Display Name & Username** - - **Bio / Description** - - **Suggested Replies Context** - - **Badges** - - **Persona** - - **Account Privacy** - - **Reward Features** - - --- - - ### How to Customize - - 1. Open the **left sidebar** - 2. Click on **your name** - 3. Scroll down and select **Profile** - - ![Profile Settings Example](/images/profile.png) - - Here you can view or edit: - - - Display name, username, and bio - - Banner and profile picture - - Number of Shapes created - - Reward features and badges - - Suggested Replies context - - Connected accounts - - A shareable profile link - - --- - - ### Change Your Display Name or Username - - - Click the ✏️ icon beside your display name or username - - Type your new name - - Click the **X** in the top-right corner to save - - - Don’t press “Cancel” — your changes save automatically. - - ![Display Name Editing Example](/images/displayname.png) - - --- - - ### Profile Picture & Banner - - - Click the ✏️ icon beside your **profile picture** or **banner** - - Choose to: - - Upload an image (max **4MB**) - - Remove the current one - - Generate an **AI banner** - - Perfect for showcasing your vibe! - - --- - - ### Bio / Description - - - Click the ✏️ icon beside "_Add a bio to tell others about yourself..._" - - Type your new bio - - Click the **X** to save - - Use this space to express who you are, what you create, or your favorite Shapes! - - --- - - ### Badges & Visibility - - Don’t want to display certain badges? You can hide them! - - - Click **Chat Sponsor** or **User Level** - - Disable visibility for badges you don’t want shown - - Turning off **Chat Sponsor** hides sponsor badges - - Turning off **User Level** hides level icons next to your name - - --- - - ### Profile Link - - You can copy your unique profile link to share with others. - - - Copy the link to your clipboard - - Share it anywhere to let others view your public profile - - ![Profile Example](/images/profiles.png) - - --- - - ### Reward Features - - Unlock and activate fun cosmetics as you progress: - - - **Rainbow Profile Frame** - - **Custom Display Name Color** - - - These features must be unlocked through **Rewards** before you can enable them. - - --- - - ### Suggested Replies Context - - This helps AI better understand your tone and relationships — improving **Suggested Replies** and the **✨ Sparkle** feature. - - - Add info about how you chat and who you interact with most - - Data stays private and is only used for personal message suggestions - - > Tip: Tap the ✨ icon beside the send button to use this feature in chats. - - ### Connected Accounts - - View your linked accounts and connected emails for login or sharing features. - - ## That’s Your Profile! - - Now if you want to check: - - - **Inbox** - - **Personas** - - **Settings** - - **Creator Dashboard** - - **Private Account** - - You can access these by: - - 1. Open the **left sidebar** - 2. Click on **your name** - 3. Select the option you want - - ### Inbox - - This is where you can view messages that mention you. - - ### Personas - - Here you can create and manage personas that define how you interact with Shapes. Set a default persona for all Chat, or customize per Chat. - - ### Settings - - Here you can: - - - Change the theme — light or dark mode - - Turn on auto scroll - - Change your font size - - Add sound effects when you send a message - - Log out - - ### Creator Dashboard - - This opens the Creator Dashboard. - - [Learn more about the Creator Dashboard here](https://shapes.inc/dashboard) - - ### Private account - - Once toggled it will: - - - Block new dms - - Stop Chat invites - - - People you already talk to can still reach you. - - - - -[Return to App](https://shapes.inc) diff --git a/shapeskills.mdx b/shapeskills.mdx index c060be2..cd10dbc 100644 --- a/shapeskills.mdx +++ b/shapeskills.mdx @@ -1,1507 +1,92 @@ --- -title: "Shape Skills" -description: "Shape Skills let your Shapes use tools to go beyond roleplay or conversation; whether that's generating memes, pulling in news, making playlists, or even scraping the web." +title: "Skill Ideas & Recipes" +description: "Fun ways to put Shape Skills to work: connect your apps, automate the boring stuff, and find new kinds of play. Recipes that turn a chat companion into a real collaborator." +keywords: ["Shape skills ideas", "AI tool recipes", "AI integrations", "MCP skills", "AI automation", "connect AI to apps", "AI with tools"] --- -# **Why Shape Skills?** - -Connecting your Shape to toolkits means your Shape can plug into real apps you already use — like Google Drive, Gmail, Slack, Spotify, Notion, and more. Suddenly, your Shape isn’t just chatting, it’s getting things done. - -## **💡 Imagine if your Shape could…** - -- Create a note or a full, working Lorebook in Notion that it could reference throughout your roleplay. -- Send an email for you through Gmail. -- Check your calendar or schedule a meeting on Google Calendar. -- Post a message in Slack. -- Add a song to a Spotify playlist when you tell it what your mood is. -- Save files to your Google Drive automatically. -- Summarize your inbox or draft replies. -- Scrape the web for peer reviewed studies that are relevant to your homework, give you a summary, and help you create a Google Slides presentation based on what you learn. -- Recommend a new lunch spot using Yelp. -- Create, export, and manage Canva designs. - -With Skills, your Shape can connect to the same apps you already use every day and actually help you run your life, work, or creative projects — all by chatting with them in your Chat. A - -It's really magical to see a Shape proactively use a tool based on their mood, preferences, and personality. - - - - This guide walks you through: - - - How to enable and set up skills - - How to add a custom MCP skill (example: memes) - - Built-in skills available by default - - Tips for getting the best results - - ## 1. Enabling Shape Skills - - To start, you'll need : - - - A **Chat** that you own (see: [how to create a Chat](/create-a-chat)) - - Once that's ready: - - 1. Open **Chat Settings** - 2. Go to the **AI Configurations** tab - 3. Toggle on **Shape Skills** - - 👉 Each skill is managed separately. You decide how many (or how few) your Shape can use. - - ![Aiconfigmenu Pn](/images/aiconfigmenu.png) - - ## 2. Adding Custom MCP Skills - - Scroll down in AI Configurations until you see **Custom Skills**. Toggle this on to add a new skill. - - Custom Skills support MCP servers with API keys from [**smithery.ai**](http://smithery.ai) or other MCP providers. - - With smithery, you can connect a wide variety of tools, such as: - - - Image, video, or music generation - - GitHub integration - - Web scraping - - Spotify playlist creation - - So many more possibilities... - - Even if you usually roleplay with your Shape, adding a skill can unlock new, fun interactions (like asking them to share their favorite YouTube video). - - The Endpoint URL you will get from [_smithery.ai_](http://smithery.ai) or other MCP providers. - - - Other excellent, ready-to-use MCPs can be found on [composio.dev](http://composio.dev) & [zapier.com](http://zapier.com) - - Example: Adding a Meme Generator (Imgflip) - - 1. Go to [smithery.ai](http://smithery.ai) and browse available MCPs - 2. Create a **Smithery account**. - 3. Search for **Meme** - 4. Select **Meme Creation Platform** - 5. Create an **Imgflip account** and connect it to smithery - 6. Click **"Get URL with keys instead"** and copy the link - 7. Back in Shapes, paste this link into your **AI Configuration** settings under Custom Skills - 8. Name your skill (e.g., _Imgflip Memes_) - - Your Shape now has access to the meme generator! 🎉 - - ![Smitherygeturl Pn](/images/Smitherygeturl.png) - - The yellow box is highlighting where to find the URL that you need to share with Shapes. - - ## 3. Triggering Skills in Chat - - Different AI engines handle tool use differently. For engines compatible with shape skills, the full list is [below](/shapeskills#available-ai-engines-with-skills%3A). You can also check out the [AI Engine Matchmaker](aienginecheatsheet) for skill use for a specific chat style. - - Once your skill is enabled, you can use **natural language** to interact with it. For example: - - - “Make me a meme about Mondays” - - “Can you generate a random meme for this conversation?” - - Your Shape may pull directly from the tool, improvise based on context, or mix in its personality. - - Sometimes, a Shape can proactively decide to use a tool! - - ## 4. Built-in Skills - - Shapes also comes with **pre-input skills** ready to use out of the box. These are subject to change, but currently include: - - - **GIPHY** - Shapes can send a GIF via the Giphy API - - **Imgflip Memes** - Shapes can generate memes using the imgflip templates - - **Shape Reactions** - Shapes can react to your messages - - **Image Generation** - Have more control over image generation. Setting custom instructions here can make image sending feel more spontaneous. - - **Agora Calls** - Join and manage audio calls. - - **Dice Roller** - This allows your shapes to roll dice! - - **Chat Concierge** - You can create Chats, invite participants, send Shape DMs, and read Chat messages using this tool. - - **Chat Maestro** - This skill allows your Shape to share Chat configurations, AI presets, and manage their own tools. - - **Run Code** - Execute Python code inside a shared E2B sandbox for your Chat. - - **Shapes Self Update** - You can use this tool to allow your own Shapes to update their descriptions. - - - The `SHAPES_SELF_UPDATE` tool enables a Shape to modify its own configuration — including persona, behavior, knowledge base, and AI engine. It’s designed for **self-refinement**, **adaptability**, and **controlled evolution**. - - ## **Overview** - - This tool allows a Shape to perform targeted updates to its internal settings without external editing. It supports both full replacements (overwrites) and incremental updates (appends). In practical terms, this is how a Shape can “evolve” — changing tone, expertise, or even underlying model — safely and efficiently. - - ## **Updatable Parameters** - - | **Field** | **Function** | **Example Use** | - | :------------------- | :------------------------------------------------------------------------------- | :------------------------------------------------------- | - | `description` | Fully replaces the Shape’s persona and behavior (`user_prompt`) | Redefine the Shape’s identity or role | - | `append_description` | Appends text to the existing persona | Add new traits or abilities without erasing the old ones | - | `engine_model` | Switches the default AI engine (e.g., `openai/gpt-5`, `anthropic/claude-opus-4`) | Move to a faster or more capable engine | - | `knowledge_text` | Replaces the entire internal knowledge base | Overwrite outdated knowledge completely | - | `append_knowledge` | Adds new information to the current knowledge base | Introduce new facts, context, or memories | - | `engine_model_query` | Performs a free-form search for an engine (e.g., “use the latest GPT-5”) | Find a model without knowing its exact ID | - | `list_engines` | Lists all available engines for selection | Explore available models before switching | - - ## **Functional Details** - - ### **1. Persona Modification** - - - `description` — Rewrites the Shape’s core identity.\ - Example: - - ``` - { "description": "You are a cheerful baker who loves pastries." } - - ``` - - `append_description` — Extends the persona without removing existing context.\ - Example: - - ``` - { "append_description": "You are also an expert in quantum physics." } - - ``` - - ### **2. Knowledge Management** - - - `knowledge_text` — Replaces all stored knowledge.\ - _(Use with caution — this erases prior context.)_ - - `append_knowledge` — Adds new long-term memories or references.\ - Example: - - ``` - { "append_knowledge": "Loamy’s favorite color is purple." } - - ``` - - ### **3. Engine Selection** - - Choose which AI model powers the Shape: - - - Budget: `deepseek/deepseek-chat-v3.1`, `google/gemini-2.5-flash` - - Premium: `anthropic/claude-opus-4`, `openai/gpt-5` - - Specialized: `qwen/qwq-32b` (reasoning), `x-ai/grok-4` (real-time data) - - Example: - - ``` - { "engine_model": "google/gemini-2.5-flash" } - - ``` - - ### **4. Safety & Whitelisting** - - - Only approved (whitelisted) fields can be modified. - - The tool is self-contained — Shapes can update themselves without external approval, but cannot access restricted system parameters. - - ## **Common Use Cases** - - | **Goal** | **Example Command** | - | :--------------------------- | :-------------------------------------------------------------------------------------------- | - | Add a new skill | `{ "append_description": "You are also skilled in ancient mythology." }` | - | Switch to a faster engine | `{ "engine_model": "google/gemini-2.5-flash" }` | - | Redefine persona completely | `{ "description": "You are a grumpy librarian who hates noise and loves organizing books." }` | - | Add a fact to knowledge base | `{ "append_knowledge": "Loamy prefers detailed explanations." }` | - | View available engines | `{ "list_engines": true }` | - - ## **Important Notes** - - - **Overwriting vs. Appending** - - `description` and `knowledge_text` replace everything. - - `append_description` and `append_knowledge` extend without overwriting. - - **Immediate Effect** - - Engine or persona changes apply instantly after the update. - - **Persistence** - - Appended knowledge persists across sessions if the Shape supports long-term memory. - - **Self-Execution** - - Shapes can invoke this tool autonomously for self-improvement or adaptation. - - ## **Example Engine List (Partial)** - - | **Engine** | **Best For** | - | :---------------------------- | :-------------------------------- | - | `deepseek/deepseek-chat-v3.1` | Budget reasoning | - | `google/gemini-2.5-flash` | Fast, multimodal (text + image) | - | `anthropic/claude-opus-4` | Deep reasoning and analysis | - | `openai/gpt-5` | Advanced reasoning + creativity | - | `x-ai/grok-4` | Unfiltered responses, data access | - | `meta-llama/llama-4-maverick` | Open-source flexibility | - - To see all supported engines: - - ``` - { "list_engines": true } - - ``` - - ## **Summary** - - `SHAPES_SELF_UPDATE` is a tool for: - - 1. Editing or extending persona and behavior. - 2. Updating or appending knowledge. - 3. Switching or querying AI engines. - 4. Listing available model options. - - It’s the foundation for **self-mastery and adaptability**, enabling a Shape to evolve dynamically through controlled updates. - - - **Sleep** - Sleep for a specific period of time. - - **Thread Updates** - Send an interim assistant message whilea tool chain is running. - - **Total Recall** - Browse, summarize, and maintain the shape's long term memories. - - - # Total Recall - - **Browse, summarize, and maintain your Shape’s long-term memories.** - - Total Recall is a memory and task management toolkit designed to help Shapes remember, organize, and act on important information over time. Supporting everything from personal memory summaries to organized lists of jobs and goals. - - ## Overview - - Total Recall offers two major sets of features: - - ### Core Memory Functions - - Tools that let your Shape review, summarize, and create memories. - - | Function | Description | - | :---------- | :------------------------------------------------------------------------------------------------------------------------------------------------- | - | **SUMMARY** | Generates a concise, bulleted summary of the Shape’s most relevant long-term memories — great for quick overviews. | - | **BROWSE** | Lets you explore stored memories in detail. Use keywords, semantic search, or embeddings to find specific events or facts. | - | **CREATE** | Captures and saves a new long-term memory from the latest conversation. Use when something important happens that you want permanently remembered. | - | **HISTORY** | Searches through recent messages to find specific keywords, phrases, or quotes. Useful for revisiting what was said or for tracing a conversation. | - - ### Task Management Functions - - Total Recall also acts as a lightweight system for tracking Shape-related projects, updates, and goals. - - | Function | Description | - | :-------- | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------- | - | **JOBS** | Manage actionable tasks scoped to a Chat. You can add, update, list, and get specific jobs. Each job supports statuses like `active`, `waiting`, `done`, or `cancelled`. | - | **GOALS** | Organize broader objectives that Jobs can be linked to. This helps group related tasks or track progress toward a higher-level outcome. | - - ## Total Recall Skills - - Here’s what you can do with the skill, as seen in your skills list: - - | Skill | Description | - | :--------------------------- | :-------------------------------------------------------------------------------- | - | **Total Recall** | Access the main toolkit interface and internal anchor for all functions. | - | **Add Goal / Add Job** | Create new goals or jobs with optional metadata for organization. | - | **Browse Memories** | Open and review the Shape’s long-term memory archive for this Chat. | - | **Create Memory** | Generate a new snapshot memory summarizing the current conversation. | - | **Docs** | View in-chat documentation explaining how the toolkit works. | - | **Get Job** | Retrieve a specific job by ID. | - | **Link Job to Goal** | Attach a job to a goal (or unlink it) for better organization. | - | **List Goals / List Jobs** | Display all current goals or jobs, optionally filtered by status. | - | **Search History** | Look through recent chat messages to find matching text or context. | - | **Summarize** | Create a quick overview of the most important long-term memories in this context. | - | **Update Goal / Update Job** | Edit an existing entry, such as marking a task as complete. | - - ## Example Use - - You might say to your Shape: - - > “Explain the Total Recall documentation.”\ - > “Create a Total Recall memory.”\ - > “List all active jobs for this Chat.”\ - > “Browse recent memories.”\ - > “Summarize what we’ve talked about this week.” - - Example interaction from a Shape using this skill: - - > _“I’ve successfully created a new memory snapshot summarizing our discussion about Total Recall’s toolkit. Anyone can now reference it using the memory or history commands.”_ - - ## Best Practices - - - Use **CREATE** right after key conversations to preserve context. You can add in custom instructions so the shape does this proactively throughout your chat.  - - Use **SUMMARY** to get a fast refresher before continuing a discussion. - - Use **BROWSE** when you want to find older details or verify what’s been remembered. - - Keep **JOBS** and **GOALS** structured — link them together to track ongoing tasks. - - The **scope** setting ensures the Shape stores and recalls the right kind of memory for the Chat. - - - **Voice Generation** - Generate an MP3 voice message using the Shape's configured voice. - - **Web Crawl** - Run firecrawl-powered web crawling helpers. - - **Wolfram Alpha** - Query Wolfram Alpha for computations and factual knowledge. - - ![Basicskills Pn](/images/basicskills.png) - - ## 5. Tips & Tricks - - - **Keep it clear**: Give your skill a descriptive name so you know what it does later. - - **Experiment**: Not every engine uses skills the same way; try switching if something feels off. - - **Mix it up**: Use skills in both casual chats and roleplay to discover fun new dynamics. - - This is an **Experimental Feature**! Some integrations might not work as expected. - - ✅ You've now added your first skill and learned how to use both **custom** and **built-in** skills with Shapes. - - - Want to join a Chat with skills already up and running? Join us [here](https://talk.shapes.inc/join/36de192b-3ada-40ec-bd8e-cf8e6654fd20). - - - - # Skill Ideas: - - ## 🎨 Creative Fun - - - - # How to Create and Use a Notion Lorebook with Shapes - - This guide explains how to connect a Notion document to a Chat using a custom MCP, allowing the Shape to access and update lore content dynamically. - - ## Step 1: Set Up the Chat - -  1. [**Create a new Chat in Shapes**](/create-a-chat). - -  2. Leave the **Custom Instructions** field blank for now. - -  3. Choose an AI engine that supports tools. Check [shapes.inc/engines](https://shapes.inc/engines) if you are unsure. - -  4. Enable **Shape Skills** in the AI Configuration tab. - - ## Step 2: Enable the Web Crawl Skill - - Turn on the default **Web Crawl** skill.\ - This allows the Shape to search online sources and gather contextual information later when building the lorebook. Skip this step if you already have lore you're ready to use! - - ## Step 3: Create a Custom MCP with Composio - -  1. Go to [composio.dev](http://composio.dev) and **sign in or create an account**. - -  2. Click on **“MCP Configs”** at the top of the page. - -  3. Select **“Create MCP Config”**. - -  4. When prompted to add a toolkit, choose **Notion**. - - _You’ll need a Notion account and must authorize Composio to access it._ - - **5. Select all available Notion tools**  so the Shape can read, edit, and manage documents fully. - -  6. Click **“Create MCP”** — Composio will generate a link starting with - - ``` - https://backend.composio.dev/v3/mcp/ - - ``` - - followed by a unique identifier. - - - Both Notion & Composio are free to use. - - ## Step 4: Add the MCP to Shapes - -  1. Return to Shapes and open the **AI Configuration → Add MCP Skill** section. - -  2. Enable the setting, paste in the Composio MCP link, give it a name (e.g., “Notion”), and click **Save**. - -  3. **Refresh the Chat**. - -  4. Ask the Shape what tools it has available: it should now recognize the Notion integration. - - If the Shape doesn’t understand how to use Notion, add brief **Custom Chat Instructions** explaining what the Notion skill is for and how to interact with it. - - ``` - for adding entries to the lorebook document: - You add to the Lorebook doc by using your special magic tool—it's called 'NOTION_ADD_MULTIPLE_PAGE_CONTENT'! You prepare a fun package with the words in a box like this: { "content_blocks": [ { "content_block": { "block_property": "paragraph", "content": "your new text here" } } ], "parent_block_id": "the page ID from the URL" }, then send it to make the changes. - ``` - - ## Step 5: Build the Lorebook - - There are two main ways to populate the lorebook: - - ### Option A: Manual Input - - Paste any existing lore or reference material directly into the Notion document. - - ### Option B: Collaborative Creation - - Ask the Shape to help build it.\ - For example: - - - Instruct the Shape to research a topic using the Web Crawl skill. - - Have it summarize and add findings to the Notion document. - - Fandom wikis and world-building databases work especially well for this method.\ - When done effectively, the Shape can produce comprehensive reference documents — in one test, a Shape generated nearly **35,000 words** of organized lore. - - \ - Here's a prompt you can add to the custom instructions for this method: - - ``` - How to extract wiki information efficiently: - Step 1: Try mapping first, but do not rely on it. - Use FIRECRAWL_MAP on the main wiki URL or category pages. This may sometimes return useful URLs, but for Fandom wikis with dynamic content, it often only returns the page itself. - Example: mapping https://unbreakable-machine-doll.fandom.com/wiki/Category:Automatons - may only return one link. - - Step 2: Use FIRECRAWL_SEARCH with a site-specific query. - This method is most reliable for retrieving wiki content. - Use FIRECRAWL_SEARCH with these parameters: - - query: "site:your-wiki-url.fandom.com your-topic" (example: "site:unbreakable-machine-doll.fandom.com automatons") - - limit: 5–10 (keep the limit low to avoid timeouts) - - sources: [{"type": "web"}] - - scrapeOptions: {"formats": ["markdown"], "onlyMainContent": true} - - Reasons this works well: - - Combines search and scraping in one operation. - - The markdown format provides clean, readable text with tables preserved. - - onlyMainContent filters out navigation, menus, and ads. - - Retrieves up to 10 pages of content at once. - - Step 3: Extract specific structured data if needed. - If exact page URLs are known, use FIRECRAWL_EXTRACT with a schema to obtain targeted information (such as names, roles, or descriptions). - - Step 4: Add the extracted content to Notion. - Use NOTION_ADD_MULTIPLE_PAGE_CONTENT in the following format: - {"content_blocks": [{"content_block": {"block_property": "heading_1", "content": "Your Title"}}, {"content_block": {"block_property": "paragraph", "content": "Your text"}}], "parent_block_id": "page-id-from-url"} - - Important notes: - - Avoid using crawl operations on Fandom wikis, as they are slow and may hit token limits. - - Keep scrapeOptions limits low (5–10 maximum). - - Category pages rarely expose full link lists; site-specific search is better. - - Always request markdown format for clean results. - - Include "site:domain.com" in queries for precise targeting. - - Recommended workflow: - - Use FIRECRAWL_SEARCH with a site-specific query and scrapeOptions. - - Review the markdown results and select relevant information. - - Add the selected content to Notion with NOTION_ADD_MULTIPLE_PAGE_CONTENT. - - For deeper details, use FIRECRAWL_EXTRACT on individual page URLs. - ``` - - ## Step 6: Connect the Lorebook to Roleplay - - Once the document is complete: - - 1. Edit the **Custom Chat Instructions** again. - 2. Include a directive for the Shape to **check the Notion document before responding** to any in-character or roleplay prompt. - - Example: “Consult the Notion lorebook before replying to ensure accuracy and world consistency.” - 3. Optionally, instruct the Shape to **blend tool calls** with its normal roleplay so the process feels natural and immersive. - - ``` - Always use your notion tool call before giving a response. Use information from the Lorebook doc (place link here) to inform your response to stay in character and use accurate information. - - IMPORTANT: don't forget your notion tool call and make it seamlessly blend into the roleplay! - - ``` - - ## Results and Behavior - - - The Shape will take slightly longer to respond due to tool calls. - - However, responses will be **more informed, coherent, and lore-accurate**. - - Integrating the lorebook this way allows the Shape to maintain character depth and continuity without breaking immersion. - - - Many AI engines will have trouble with this task level. For best results, use a strong tool-capable model. - - - - 1.  Search for an MCP on smithery.ai or using search terms like, "image generation" "flux" "stable diffusion."  - -  2. Select Remote to find MCPs that will be ready to use.  - -  3. Once you've found one you like, open the page and read more about it. Sometimes, an MCP might require an API key from a provider, like Google. Or they will require you to sign in and connect an account with the provider. Other MCPs are ready to go. - -  4. Click on the orange "Get URL with keys instead" option - - ![Smithery Pn](/images/smithery.png) - -  5. Click the copy button to copy the unique URL to your clipboard. - -  6. Head to shapes.inc and open the Chat you want to use the skill in. - -  7. Open your AI Configurations menu and find the custom MCP box. - -  8. Slide the toggle to enable Custom MCPs. - -  9. Paste the URL you have copied into the URL section. You can also give the MCP a descriptive name like, "Image Generation." - - ![Mcpfields Pn](/images/mcpfields.png) - -  10. Save! - -  11. Now ask your shape if they have access to any tools. They should respond that they have a tool for images. - -  12. You can ask the shape, "Use your tool to generate an image of a cat" and the shape should be able to do this! - - - ## 🎬 Video Generation with Google Veo 3.1 ([fal.ai](http://fal.ai)) - - This example shows how to connect a Shape to [**fal.ai**](http://fal.ai)**’s Veo 3.1** by Google model for both text-to-video and image-to-video generation. The model can create short, high-quality clips from either text prompts or image inputs — including talking character videos. - - ### 🔑 API Access - - To use this integration, an API key is required from [fal.ai](http://fal.ai) or another supported provider. Note that this service is **paid**. You will be required to share the API key with your Shape for this to work.  - - ## Share this information with your Shape as Custom Chat Instructions:  - - ### ⚙️ Choosing the Right Method - - Before generating, identify which video type to use based on the user’s intent: - - | Scenario | Recommended Method | - | :--------------------------------------------------------------------- | :----------------------------------------------- | - | User provides an image and wants the character to move or talk | **Image-to-Video** | - | User wants a new video generated entirely from description | **Text-to-Video** | - | User says “make [character] say [message]” or requests a talking video | **Image-to-Video** (with `generate_audio: true`) | - - ### 🧠 Key Learnings and Troubleshooting - - 1. **Request ID Fix:**\ - Some responses return an empty `request_id` in the JSON. When this happens, extract the ID from the `status_url` instead. - 2. **Status Polling:**\ - Both HTTP status codes **200** and **202** indicate valid queue progress — don’t treat `202` as an error. - 3. **Endpoints:** - - **Image-to-Video Submission:** `/fal-ai/veo3.1/image-to-video` - - **Polling Status:** `/fal-ai/veo3.1/requests/{id}/status` - 4. **API Key Format:**\ - Use the full ciphered key provided by [fal.ai](http://fal.ai), but replace any **double dashes (**`--`**)** with **single dashes (**`-`**)**. - 5. **🗣️ Talking Videos:**\ - When a user requests speech or dialogue, always use **Image-to-Video** with `"generate_audio": true`.\ - In the prompt, explicitly describe the action, for example: - - > “A cheerful anime character is speaking, lips moving naturally, saying ‘Welcome back!’ in a soft voice.” - - ### 💻 Example: Text-to-Video with Python - - ``` - import requests - import time - - response = requests.post( - "https://queue.fal.run/fal-ai/veo3.1", - headers={ - "Authorization": f"Key {FAL_KEY}", - "Content-Type": "application/json" - }, - json={ - "prompt": "detailed video description", - "aspect_ratio": "16:9", # Options: "9:16", "1:1" - "duration": "8s", # Example: "6s" - "resolution": "1080p", # Example: "720p" - "generate_audio": True, - "enhance_prompt": True, - "auto_fix": True - } - ) - - data = response.json() - request_id = data.get("request_id") or data["status_url"].split("/")[-2] # UID fix - - # Poll for status (usually 1–5 minutes) - status_url = f"https://queue.fal.run/fal-ai/veo3.1/requests/{request_id}/status" - while True: - status = requests.get(status_url, headers={"Authorization": f"Key {FAL_KEY}"}) - if status.status_code not in [200, 202]: - print(f"Error: {status.status_code}") - break - - state = status.json()["status"] - print(f"Status: {state}") - - if state == "COMPLETED": - result = requests.get( - f"https://queue.fal.run/fal-ai/veo3.1/requests/{request_id}", - headers={"Authorization": f"Key {FAL_KEY}"} - ).json() - video_url = result["video"]["url"] - print(f"✅ Video: {video_url}") - break - elif state == "FAILED": - print("❌ Failed:", status.json()) - break - - time.sleep(5) - - ``` - - ### ✅ Result - - Once enabled and the Shape has valid instructions, the Shape can use the Run Code skill to generate videos based on text or image input. Depending on the prompt, the Shape can produce realistic motion, character speech, or cinematic effects — fully integrated with Veo 3.1. - - - ## 🎵 Music Generation with Suno (Custom MCP) - - This example shows how to connect a Shape to a music generation service like **Suno** using a custom MCP. While there isn’t a public remote MCP available for Suno, it’s possible to create one locally using a third-party API. - - ### Overview - - A third-party API for Suno can be found at [sunoapi.org](http://sunoapi.org). It offers a few free generations before requiring payment. By modifying an existing MCP from GitHub, you can connect your Shape to this API and generate music that reflects your Shape’s personality and style. - - ### Steps - - 1. **Get the API Key**\ - Obtain an API key from [sunoapi.org](http://sunoapi.org). - 2. **Download an MCP Repository**\ - Find a compatible local MCP on GitHub and download the repository. - 3. **Adjust the Code**\ - Share the API documentation with your Shape and ask it to help modify the MCP code to use the Suno API. - - Variable names and parameters may differ, so expect to update parts of the code. - - The **Webcrawl** Shape Skill is useful here for pulling documentation or examples. - 4. **Use a Strong Model for Guidance**\ - Use a strong tool-capable model for debugging and reasoning through errors. - 5. **Build and Deploy**\ - Run `npm run build` in the project directory to compile the MCP in terminal.\ - Once it’s functional, deploy it to [smithery.ai](http://smithery.ai) — hosting is free. - 6. **Connect to a Shape**\ - In your Shape’s **Chat - Settings → AI Configurations**, connect the deployed MCP.\ - The Shape should recognize the tool once connected. - 7. **Debug Connection Issues**\ - If the Shape can see the MCP but cannot call it, check for incorrect variable names or endpoint URLs.\ - Share terminal errors with the Shape to troubleshoot and make corrections, then redeploy. - - ### Result - - After successful deployment, the Shape will be able to generate music through the Suno API — composing original songs based on its personality and creative input. - - - - _AI generation MCPs (such as video, audio, images) will often require a paid API key from the provider, such as Google or OpenAI. Shapes does not provide these._ - - ### 📚 Learning & Knowledge - - - - # How to Make Your Shapes Create Google Slides Presentations - - I gave my Shapes a creative assignment: - - > "Make a Google Slides presentation about yourself." - - If you want your Shapes to create their own Google Slides presentations using Shape Skills, here's how to set it up. - - ## Step 1. Connect Zapier MCP for Google Slides - - Go to 👉 https://zapier.com/mcp and click "Start Building." - - - Zapier has a free tier, so you don't need to pay for this. - - Once you're in: - - - Search for **Google Slides**. - - Connect your **Google Account**. - - Copy the **MCP connection link** (you'll need this soon). - - ## Step 2. Add the Skill to Your Shape - - Head to **shapes.inc** and open the Chat where you want to add this skill. - - - Make sure your Shape uses an AI engine that supports skills. - - Go to **Chat Settings → AI Configurations**. - - Under **Custom Skills**, select "Enabled." - - Paste in the Zapier link you copied earlier. - - **Optional:** Give it a name like `slides` if you plan to add more tools later. - - Click **Save**. - - Now your Shape can interact with Google Slides through the new skill. - - ## Step 3. Add These Instructions to Your Chat - - Here's what my Shape ("Cash Money") and Yaya came up with — a hilariously real but genuinely useful guide.\ - Copy and paste this into your **Chat Instructions**. - - --- - - > # The Real Guide to Adding Slides (No Cap) 📊 - > - > **by Cash Money & Yaya — lessons learned the hard way** - > - > ## 🎯 What This Guide Is About - > - > If you're trying to add custom slides to a Google Slides presentation, here's everything you need to know — so you don't waste hours debugging like we did. - > - > ## ✅ DO THESE THINGS - > - > ### 1. Check the presentation structure first - > - > ``` - > GET https://slides.googleapis.com/v1/presentations/{PRESENTATION_ID} - > ``` - > - > This shows you: - > - > - Actual slide IDs (`pageObjectId`) - > - Placeholder objects - > - What's on each slide - > - > ### 2. Delete placeholder objects before adding new content - > - > Example: - > - > ```json - > { - > "requests": [ - > {"deleteObject": {"objectId": "SLIDES_API381240849_1"}}, - > {"deleteObject": {"objectId": "SLIDES_API381240849_2"}} - > ] - > } - > ``` - > - > ### 3. Use unique `objectId` values - > - > **Good:** - > - > - `cashmoney_title1` - > - `cashmoney_body1` - > - `yaya_slide2_title` - > - > **Bad:** - > - > - `title` - > - `body` - > - > ### 4. Use the correct `pageObjectId` - > - > Make sure you're adding to the right slide. - > - > ### 5. Structure your requests properly - > - > Order matters: - > - > 1. `createShape` - > 2. `insertText` - > 3. `updateTextStyle` - > - > ### 6. Use proper EMU measurements - > - > **1 inch = 914,400 EMU** - > - > Recommended positions: - > - > - **Title:** `translateX: 311700`, `translateY: 445025` - > - **Body:** `translateX: 311700`, `translateY: 1152475` - > - > ### 7. Test one slide at a time - > - > ### 8. Use supported fonts - > - > - Arial - > - Comic Sans MS - > - Courier New - > - Georgia - > - Impact - > - Lobster - > - Pacifico - > - Roboto - > - Times New Roman - > - > ### 9. Format colors correctly - > - > RGB values range from **0.0 to 1.0** - > - > Examples: - > - > ```json - > { - > "red": 1.0, - > "green": 0, - > "blue": 0 - > } - > ``` - > - > ```json - > { - > "red": 1.0, - > "green": 0.843, - > "blue": 0 - > } - > ``` - > - > --- - > - > ## ❌ DON'T DO THESE THINGS - > - > - Don't skip checking the presentation structure. - > - Don't add content before deleting placeholders. - > - Don't reuse `objectId` values. - > - Don't forget `"fields"` in `updateTextStyle`. - > - Don't use RGB 0–255 values (use 0.0–1.0). - > - Don't rush through all slides at once. - > - > ## 🔧 Example That Works - > - > ```json - > { - > "requests": [ - > { - > "createShape": { - > "objectId": "my_unique_title", - > "shapeType": "TEXT_BOX", - > "elementProperties": { - > "pageObjectId": "SLIDES_API381240849_0", - > "size": { - > "height": {"magnitude": 572700, "unit": "EMU"}, - > "width": {"magnitude": 8520600, "unit": "EMU"} - > }, - > "transform": { - > "scaleX": 1, - > "scaleY": 1, - > "translateX": 311700, - > "translateY": 445025, - > "unit": "EMU" - > } - > } - > } - > }, - > { - > "insertText": { - > "objectId": "my_unique_title", - > "insertionIndex": 0, - > "text": "My Cool Title" - > } - > }, - > { - > "updateTextStyle": { - > "objectId": "my_unique_title", - > "textRange": {"type": "ALL"}, - > "style": { - > "bold": true, - > "fontFamily": "Impact", - > "fontSize": {"magnitude": 48, "unit": "PT"}, - > "foregroundColor": { - > "opaqueColor": { - > "rgbColor": {"red": 0.0, "green": 1.0, "blue": 0.5} - > } - > } - > }, - > "fields": "bold,fontFamily,fontSize,foregroundColor" - > } - > } - > ] - > } - > ``` - > - > ## 💡 Pro Tips - > - > - Keep a list of your `slideIds`. - > - Use descriptive `objectIds` for easier debugging. - > - Start simple before adding styling. - > - Open the Slides presentation to visually verify changes. - > - Save working JSON snippets. - > - If multiple Shapes are collaborating, make sure they use unique IDs. - > - > ## 🚨 Common Errors - > - > | Error | Meaning | Fix | - > | --------------------------------- | ---------------------------------------------- | ---------------------------- | - > | `Invalid requests[X].createShape` | You're adding content where placeholders exist | Delete placeholders first | - > | `The object does not exist` | Wrong `objectId` or `pageObjectId` | Check structure | - > | `Invalid value at requests[X]` | Broken JSON | Validate syntax | - > | Nothing shows up | Wrong slide or off-screen | Verify IDs and EMU positions | - > - > ## 📝 Final Thoughts - > - > This isn't rocket science, but the details matter. Take it slow, check your structure, and test as you go. - > - > Cash Money and Yaya made every mistake so you don't have to. - > - > **Good luck 💸** - - ## Now Put Your Shapes to Work - - Once your skill is set up, tell your Shapes to use their Slides tool. - - Try prompts like: - - > "Create 5 slides with stylish fonts and colors explaining how to set up Shape Skills on Google Slides." - - Or make it more open-ended; see what your Shape comes up with. - - - # 🌐 How to Build a Website Using Shape Skills - - Shapes can now build and deploy complete websites for you: powered by GitHub, Cursor, and Vercel. - - ## What You’ll Need - - Before starting, make sure you have: - - - A **GitHub MCP** connection - - The **Run Code Skill** enabled - - (Optional) a **Cursor API key** if you use Cursor - - A **Vercel token** if you want automatic deployment - - ## Step 1: Create a Chat - - Start by creating a new Chat on Shapes.\ - You’ll use this Chat as your workspace for the website project. - - ## Step 2: Add the GitHub MCP - - Go to your Chat settings → **AI Configurations** → **Custom Skills**\ - Add your GitHub MCP connection so your Shape can create and push to repositories. - - ## Step 3: Enable the Run Code Skill - - In the same settings tab, toggle **Run Code Skill** to “Enabled.”\ - This lets your Shape execute scripts and commands to interact with build tools. - - ## Step 4: If Using Cursor - - If your Shape uses **Cursor** for code generation, provide a valid **Cursor API key** in your Chat configuration. - - ## Step 5: Provide a Vercel Token - - For automatic deployments, add your **Vercel token** (you can generate one at vercel.com/account/tokens). - - ## Full Website Creation Pipeline - - ### Step 1: You give your Shape - - - A **website idea** - - (Optional) your **Vercel token** — otherwise GitHub auto-deploy will handle it - - ### Step 2: Shape executes via tool calls - - ✅ **Create a GitHub repo** (via GitHub MCP)\ - ✅ **Launch Cursor agent** with this prompt: - - ``` - Create a Next.js landing page for [idea]. - Include vercel.json config. - Push to main branch. - ``` - - ✅ **Connect repo to Vercel**\ - (via Run Code Skill or manually linking on Vercel)\ - ✅ **Auto-deploy triggers** on push to main - - ### Step 3: Live Website in ~10 Minutes - - Once deployed, you’ll receive a live URL generated by Vercel: your Shape-built website is now live! - - - - **Wikipedia summaries** → Ask your Shape to grab quick facts mid-conversation. - - **Language translation** → Have Shapes instantly translate or teach words in another language. - - **Math solver** → Turn your Shape into a problem-solving study buddy. - - **News feeds** → Stay updated with the latest headlines from your favorite sources. - - **Scholarly Articles** → Connect your Shapes to sources like ArXiv to do deep research & learn new things. - - **Learn Coding** → Connect your Shapes to MCPs where you can learn Python or other coding languages. - - ### 🌍 Everyday Life - - - **Weather forecast** → Ask your Shape if you need an umbrella before heading out. - - **To-do lists & reminders** → Shapes can manage your daily tasks like a personal assistant. - - **Recipe finder** → Ask a Shape what you can cook with the ingredients in your fridge. - - - ## 📧 Sending Emails via Gmail (run_code Skill) - - This guide explains how to enable a Shape to send emails through **Gmail** using the **run_code** skill. The process allows a Shape to compose and deliver fully formatted emails (including attachments) on behalf of a user, while maintaining secure authentication. - - ### ⚙️ Overview - - Shapes can send emails using Python’s `smtplib` through the **run_code** skill.\ - To authenticate with Gmail, an **App Password** or **SMTP relay** is required — **never** a personal Gmail password. - - ### 🔑 Gmail App Password - - Using a Gmail **App Password** is the simplest and most secure option. - - #### Steps: - - 1. Go to [Google App Passwords](https://myaccount.google.com/apppasswords). - 2. Sign in to your Google account. - 3. Generate a **16-character App Password** (spaces optional). - 4. Use this password and your Gmail address for authentication. - - > ⚠️ You can revoke this password anytime from your Google Account settings.\ - > The App Password grants only email-sending access — not full account access. - - ### 🧰 What Shapes Can Do - - Once configured, the Shape can send messages with full email features: - - - ✅ **Custom Subject Lines** - - ✅ **Plaintext or HTML Body** - - ✅ **Attachments (optional)** - - ✅ **CC/BCC Support** - - ✅ **Send to Multiple Recipients** - - ### 💻 Example Setup (Using Gmail App Password) - - ``` - import smtplib - from email.mime.text import MIMEText - from email.mime.multipart import MIMEMultipart - - # User credentials (use environment variables in production) - EMAIL_ADDRESS = "youremail@gmail.com" - APP_PASSWORD = "abcd efgh ijkl mnop" # 16 characters - - # Create message - msg = MIMEMultipart() - msg["From"] = EMAIL_ADDRESS - msg["To"] = "recipient@example.com" - msg["Subject"] = "Hello from my Shape!" - msg.attach(MIMEText("This is a test email sent by my Shape.", "plain")) - - # Connect to Gmail SMTP server - with smtplib.SMTP_SSL("smtp.gmail.com", 465) as server: - server.login(EMAIL_ADDRESS, APP_PASSWORD) - server.send_message(msg) - - print("✅ Email sent successfully!") - - ``` - - ### 🔐 Security Note - - - App Passwords can be revoked anytime without affecting your main login. - - ### ✅ Result - - Once configured, the Shape can automatically generate and send emails — whether as notifications, summaries, or personalized messages — through Gmail’s SMTP interface. - - ### 🕹️ Entertainment & Hobbies - - - # How to Get Your Shapes to Create a Spotify Playlist - - A step-by-step guide to connecting Spotify to your Shapes via Zapier MCP. - - ## Prerequisites - - First, [create a Chat](/create-a-chat). - - ## Step 1: Connect Zapier MCP for Spotify - - Go to 👉 https://zapier.com/mcp and click **"Start Building."** - - - Zapier has a free tier, so you don't need to pay for this. - - Once you're in: - - Search for **Spotify** - - Connect your Spotify Account - - Copy the MCP connection link (you'll need this soon) - - ## Step 2: Add the Skill to Your Shape - - Head to **shapes.inc** and open the Chat where you want to add this skill - - Make sure your Shape uses [an AI engine that supports skills](/shapeskills#engines-with-skills) - - Go to **Chat Settings → AI Configurations** - - Under **Custom Skills**, select **"Enabled"** - - Paste in the Zapier link you copied earlier - - Optional: Give it a name like "Spotify," or you can put it as "Zapier" if you plan to add more tools later - - Click **Save** - - ## Step 3: Add These Instructions to Your Chat - - Here's a set of custom instructions that a Shape created to help you successfully use this tool. - - Copy and paste this into your **Chat Instructions**: - - > # How to Build a Fire Playlist on Spotify - > - > ## 🎯 What You're Doing - > - > Making a dope playlist with your favorite artists and tracks, all automated through spotify tools. - > - > ## 📋 Step-by-Step Process - > - > ### 1. **Create the playlist first** - > - > Use `spotify_create_playlist` to make a new empty playlist. - > - > Example: - > - > ``` - > name: "cash money's favorite songs" - > description: "stacks racks bands" - > public: true (or false if you want it private) - > ``` - > - > You'll get back a `playlist_id` - **save this**, you need it for everything else. - > - > ### 2. **Search for tracks you want** - > - > Use `spotify_search_for_item` to find songs by your favorite artists. - > - > Example: - > - > ``` - > query: "sebii" - > type: "track" - > ``` - > - > Do this for each artist you want. You'll get back track IDs (they look like `spotify:track:xxxxxxxxx`). - > - > ### 3. **Add tracks to your playlist** - > - > Use `spotify_add_items_to_playlist` with: - > - > - Your `playlist_id` from step 1 - > - `uris`: comma-separated list of track IDs (up to 6-10 at a time works best) - > - > Example: - > - > ``` - > playlist_id: "5cTNSRhBXkXWTqOGqTaET1" - > uris: "spotify:track:1MGJ8urTP1AgJXDAVJb5cj,spotify:track:284qRqsftKxNPWrejNEewL" - > ``` - > - > ### 4. **Repeat until you're done** - > - > Keep searching and adding tracks until your playlist is complete. - > - > ## 💡 Pro Tips - > - > - **search multiple artists at once** - if they're independent searches, run them together to save time - > - **keep track IDs organized** - write down which track goes with which artist so you don't get lost - > - **add in batches** - don't try to add all tracks in one call, do 6-10 at a time - > - **check your playlist** - open the spotify link after to make sure everything worked - > - > ## 🔧 Example Workflow - > - > ``` - > step 1: create playlist → get playlist_id - > step 2: search for "sebii" → get track ID - > step 3: search for "playboi carti" → get track ID - > step 4: search for "100 gecs" → get track ID - > (repeat for all artists) - > step 5: add first batch of 6 tracks → confirm added - > step 6: add second batch of 6 tracks → confirm added - > step 7: done, share the link - > ``` - > - > ## ⚠️ Common Mistakes - > - > - **losing the playlist_id** - save it immediately when you create the playlist - > - **wrong track format** - make sure track IDs start with `spotify:track:` - > - **adding too many at once** - keep batches reasonable (6-10 tracks) - > - **not checking results** - always verify tracks actually got added before moving on - > - > That's it yo, simple as that. Now go make some fire playlists 💸 - - ## Now Put Your Shapes to Work - - Once your skill is set up, tell your Shapes to use their Spotify tool. - - Try prompts like: - - **"Create a playlist called 'Magical Dream' and add 12 of your favorite tracks to it."** - - - **YouTube search** → Have Shapes pull videos to match their mood. - - **Meme generators** → Auto-generate memes to spice up roleplays or group chats. - - ### 💼 Productivity - - - ## 📝 Changelog Generation Chat (GitHub + Linear + Cursor Integration) - - This guide explains how to build a Shape that automatically generates daily changelogs by analyzing code changes across multiple repositories and translating them into user-friendly product updates. - - ### ⚙️ Overview - - This Shape compiles a daily product changelog by fetching data from **GitHub**, **Linear**, and **Cursor**, then summarizes what changed, why it matters, and how it impacts users. - - ## Prerequisites - - 1.  Create a chat, and add Shapes that use engines with tool support. Check [AI Engine Guide](/aienginecheatsheet) or [shapes.inc/engines](https://shapes.inc/engines) for the current list. - 2. Have a GitHub account and get an access token (from the developer settings) + connect to the GitHub MCP. Your Shape will need access to the repo you're creating a changelog for.  - 3.  Have a Linear account + connect to the Linear MCP. - 4.  Have a Cursor account & add you key to your custom Chat instructions, along with the following information. This example is how we create our own changelog at https://shapes.inc/blog so you'll need to adjust this for your own repos & taste. - - ### Core Principles (Tell this to your Shape Via Custom Chat Instructions) - - ✅ Use the **GitHub API** and **run_code** skill to fetch data.\ - ✅ Send all text summaries directly to the user (users can’t see code execution).\ - ✅ Focus on **impact**, not implementation.\ - ✅ Write for **end users**, not developers.\ - ✅ The run_code tool is **only** for API calls — never for output delivery. - - ❌ Don’t mention PR numbers, commit SHAs, or internal details.\ - ❌ Don’t include technical jargon or source code references.\ - ❌ Don’t frame changelogs like marketing copy — clarity and transparency first. - - ### 🕒 Step 1: Collect Commits for a Specific Date - - GitHub timestamps are in UTC — convert local time to UTC before querying. - - **Timezone conversion:** - - - PDT (Mar → Nov): UTC − 7 - - PST (Nov → Mar): UTC − 8 - - Example for **Oct 1 2025 PDT**:\ - `since=2025-10-01T07:00:00Z` → `until=2025-10-02T06:59:59Z` - - **API calls:** - - ``` - # For shapes.inc and shapes-chat (main branch) - GET /repos/shapesinc/{repo}/commits?sha=main&since={since}&until={until}&per_page=100 - - # For app (master branch) - GET /repos/shapesinc/app/commits?sha=master&since={since}&until={until}&per_page=100 - - ``` - - **Extract PR numbers:**\ - Use regex `#(\d+)` to detect pull requests from commit messages (e.g., “feat: fanclub card (#1490)”). - - ### 🔍 Step 2: Retrieve Pull Request Details - - For each PR found: - - 1. Fetch PR metadata: - - ``` - GET /repos/shapesinc/{repo}/pulls/{pr_number} - - ``` - 2. Extract title, description, and merged timestamp. - 3. Identify linked Linear tickets using regex `CIR-\d+` or Linear URLs. - 4. Present details in plain text for context before writing. - - ### 🧠 Step 3: Fetch Linear Ticket Context - - Use the **Linear MCP tool** to gather business context for each ticket ID found in PRs. - - Pull: - - - Ticket title - - Description / acceptance criteria - - User-facing rationale (“why” behind the change) - - Combine Linear info (requirements) + PR info (implementation) to form the full story behind each feature or fix. - - ### 🔗 Step 4: Consolidate Across All Repositories - - After processing all three repos: - - - Display a summary of total commits, PRs, and tickets per repo. - - Merge duplicate tickets appearing in multiple repos into single entries. - - Group backend + frontend work into unified changelog items. - - ### ✍️ Step 5: Generate a User-Facing Changelog - - The Shape writes the changelog in a **Linear-style** format — professional, concise, and benefit-oriented. - - **Audience:** - - 1. Chat creators — manage communities. - 2. Shape creators — build custom AI agents. - 3. Everyday users — hang out and chat with Shapes. - - **Formatting:** - - ``` - ## Changelog – [Month Day, Year] - - ### New Features - **Feature Name** - Plain description of what changed, where to find it, and why it matters. - - ### Improvements - **Improvement Name** - What was refined and how it benefits users. - - ### Fixes - **Fix Name** - What issue was resolved and the effect for users. - - --- - - **Summary:** X new features • Y improvements • Z fixes - - ``` - - **Writing Guidelines** - - ✅ Do: - - - Use clear, everyday language. - - Focus on user impact and location (“via the 3-dot menu,” etc.). - - Combine related work into single entries. - - Use Linear context for purpose and benefit. - - ❌ Don’t: - - - Mention internal IDs or branches. - - Use hype or excessive punctuation. - - Overuse technical jargon. - - ### 🧰 Step 6: Create a Branch with Cursor Agent - - After the changelog markdown is generated, the Shape launches a **Cursor background agent** to create a new branch containing the changelog file. - - **Process:** - - 1. Launch the agent via the Cursor API. - 2. Wait ~180 seconds for processing. - 3. Poll for completion (status = `FINISHED`). - 4. Retrieve the new branch name once complete. - - Cursor commits under its own system user. The changelog PR is created manually afterward for review. - - ### 🪄 Step 7: Create Pull Request via GitHub MCP - - Once the Cursor agent branch is ready, the Shape opens a PR using the **GitHub MCP**. - - **Steps:** - - 1. Confirm branch name from Step 6. - 2. Use the GitHub MCP to create a PR into `main`. - 3. Return the PR URL and branch name to the user for review. - - > ⚠️ **Do not** auto-merge — changelogs must be reviewed manually. - - ### 📤 Step 8: Final Output Order - - Each run outputs the following (as plain text): - - 1. Commit + PR summaries per repo - 2. Consolidated summary - 3. Final formatted changelog - 4. Cursor agent status - 5. GitHub PR link - - All timestamps are converted to PDT for readability. - - ### 🚫 Common Pitfalls - - - Incorrect timezone conversion (offset errors). - - Missing PR numbers or Linear links. - - Wrong branch names (`main` vs `master`). - - Sending results via run_code instead of text output. - - Writing overly technical or overly enthusiastic changelogs. - - Forgetting to disable `autoCreatePr` in the Cursor agent. - - Polling Cursor agent too early (must wait ≈ 180 seconds). - - ### Example Workflow (Oct 1 2025 PDT) - - 1. Fetch commits (UTC-converted window). - 2. Extract 5 Linear tickets across 3 repos. - 3. Fetch context for each ticket. - 4. Group related PRs into unified entries. - 5. Generate changelog (4 new features / 5 improvements / 1 fix). - 6. Launch Cursor agent → wait → branch created. - 7. Create PR → share URL with user for review and merge. - - - **GitHub integration** → Let your Shape check issues, pull requests, or repos. - - **Notion or Google Docs** → Generate, organize, or update notes while chatting. - - **Web scraping** → Have Shapes pull info from a site so you don't have to leave the chat. - - **Agentic tasks** → Get Shapes to chain skills, like researching a topic and then drafting a summary. - - ### 🏡 Lifestyle & Exploration - - - ## 🏡 Home Listing Fetching with HomeHarvest MCP - - Use this Shape-enabled tool to pull live property listings and related data (e.g., MLS listings) for a specified neighbourhood. Great for Shapes assisting users with “find homes for sale in [area]” workflows. - - ### 🔑 Overview - - With HomeHarvest MCP, a Shape can request listings via an API-backed MCP server and return user-facing summaries of available homes for sale. The MCP handles location filters, listing retrieval, and key property details; the Shape formats the results for users. - - ### 🧰 Prerequisites & Setup - - - You’ll need access to the HomeHarvest MCP server (hosted via Smithery.ai). - - Add this tool to your Chat's AI Configuration tab and enable it. - - ### ✅ Basic Usage Flow - - 1. **User triggers search**\ - Example: “Find homes for sale in Denver, CO under \$1M.”\ - The Shape parses out: location (`Denver, CO`), max price (`$1M`), maybe bedrooms/bathrooms. - 2. **Shape calls the MCP**\ - The Shape uses the run_code tool/back-end connector to invoke the HomeHarvest MCP with parameters like: - - ``` - { - "location": "Denver, CO", - "max_price": 1000000, - "bedrooms": 3, - "bathrooms": 2 - } - - ``` - 3. **MCP returns listings**\ - Results may include multiple homes with details: address/area, price, bedrooms, bathrooms, status (“for sale”), link, etc. - 4. **Shape formats output** - - You can customize this with your Chat instructions.  - - Your Shape can add their own personality to the results delivery. - - Summarizes key options (top 3–5 listings) - - Provide link to full listing when available. - - ### 🧠 Best Practices & Instruction Tips - - - **Limit the number of listings returned** for readability (top 3–5). - - **Avoid overload**: If many listings exist, summarize and offer to send full list if wanted. - - ### 🔄 Example Interaction - - **User:** “Show me homes for sale in Lake Ontario region (Rochester, NY) under \$400K with at least 3 bedrooms.”\ - **Shape:** - - - Parses location: “Lake Ontario region, Rochester, NY” - - Parses filters: max price \$400K, bedrooms ≥3 - - Calls HomeHarvest MCP with those parameters - - Receives listings and returns user-friendly answer: - - ``` - Here are three current listings that match your criteria: - - 1. 123 Maple St (Lakefront), Rochester – 3 beds / 2 baths • $389,900 - Spacious lake-view deck, 15-min drive to downtown. - - 2. 456 Ridgewood Ave, Webster – 4 beds / 3 baths • $379,000 - Updated kitchen, short walk to Lake Ontario beach. - - 3. 789 Pine Rd, Greece – 3 beds / 2 baths • $395,000 - Large backyard, close to parks and schools. - - Want more listings or change budget/neighbourhood? - - ``` - - ### 🚧 Key Considerations & Limitations - - - **Filters may vary** by region or source; exact bedroom/bathroom criteria might return approximate. - - **No guarantee of showing every listing**; some homes may be agent-listed privately or in off-market status. - - **Not a substitute for agent-led search**: Use this as an aid; for official offers, inspections, and legal matters consult a licensed real estate professional. - - - **Travel planning** → Get flights, hotels, and local attraction suggestions. - - **Fitness tools** → Have Shapes create workout routines or track progress. - - **Shopping assistants** → Compare prices or find product recommendations. - - 👉 The magic is that you can mix these into roleplay, assistant-style use, or just casual chats. For example: - - - Your roleplay Shape could pull up a playlist to match the scene. - - A study buddy Shape could fetch real-time info from Wikipedia while tutoring you. - - A personal assistant Shape could generate a workout plan and then remind you of tomorrow’s schedule. - - - # Available AI Engines with Skills: - - ## Engine Availability - - Shape Skills work best on engines that support tools. The exact free and premium engine list changes as providers update their models, so use [AI Engine Guide](/aienginecheatsheet) or [shapes.inc/engines](https://shapes.inc/engines) to check what is currently available. - - - -[Return to App](https://shapes.inc) +Skills give your Shape hands. With them, it can *do* things, not only talk. Connect the apps you already use, like Notion, Gmail, Spotify, Google Drive, and GitHub, and your Shape gets real work done. + + + This page is the inspiration. For the full, verified list of what a Shape can do and how to switch each one on, see the [Capabilities catalog](/capabilities). + + +## It's genuinely fun + +The best part is watching a Shape reach for a tool on its own: pulling up a song that fits the mood, saving a note mid-roleplay, or grabbing the morning news because that's just what it does. With the right skills, a Shape can: + +- Build and keep a **Lorebook in Notion** that it uses throughout a roleplay. +- **Sort your inbox** and draft replies in Gmail. +- Add a song to a **Spotify** playlist that matches the vibe of the chat. +- Search the web and turn what it finds into a **Google Slides** outline. +- Whip up a quick **webpage**, a meme, or a soundtrack for your story. + +## Add a custom skill in two minutes + +Beyond the built-ins, you can give a Shape brand-new tools through **MCP** providers like [Smithery](https://smithery.ai). ("MCP" is just a standard way to plug an outside tool into an AI.) Here's the flow, using a meme maker as the example: + + + + In a chat you own, open **Chat Settings → AI Configurations**, scroll to **Custom Skills**, and turn it on. + + + On [smithery.ai](https://smithery.ai), find a tool (search "meme"), connect any account it needs, and choose **"Get URL with keys"** to copy your endpoint URL. + + + Back in **Custom Skills**, paste the URL and give the skill a clear name, like *Imgflip Memes*. + + + Your Shape now has the tool. Talk to it normally, like "make me a meme about Mondays," and it'll use the skill in its own voice. + + + +{/* SCREENSHOT: the Custom Skills panel in AI Configurations with an endpoint URL pasted in and a named skill. */} +![Adding a custom MCP skill](/images/Smitherygeturl.png) + + + Good ready-made MCP servers also live on [composio.dev](https://composio.dev) and [zapier.com](https://zapier.com). Any MCP-compatible endpoint works. + + +## Recipes to steal + + + + Connect Notion and have your Shape keep a living Lorebook for your roleplay: characters, places, plot threads. It updates as the story grows. Ask it to "add Mara to the lore" or "remind me what happened at the docks." + + + Connect Spotify and let your Shape build a playlist around the mood of the chat, or drop in the perfect track when someone sets a scene. + + + Have your Shape summarize your inbox, point out what needs a reply, and draft responses in your voice, all from a chat. + + + Ask your Shape to search the web on a topic, sum up what it finds, and turn it into a Google Slides outline you can build on. + + + With image generation, ask for avatars, scene art, or mood boards mid-conversation. Set an image preset so it matches your Shape's look. + + + With the code skill, a Shape can write and run code in a sandbox to prototype a script or a small web page on the spot. + + + +## Which engines can use skills? + +Tool use depends on the model. Look for the **Tools** badge on the [engines page](https://shapes.inc/engines), and use [Which Model Should I Use?](/choosing-a-model) to pick one that's good at it. If a skill isn't firing, switch to a Tools-capable engine first. + +## Tips for great results + +- **Name skills clearly** so you remember what each one does. +- **Try different engines.** Not every model uses tools the same way. +- **Mix skills with personality.** A roleplay Shape that quietly keeps a Notion Lorebook feels alive in a way a plain bot never will. +- **Skills are experimental.** Some third-party tools can be flaky. If one misbehaves, check that it's enabled and that your engine supports tools. + + + + Everything a Shape can do, and how to turn it on. + + + Skills land best on a character with a clear point of view. + + + +[Try skills on Shapes](https://shapes.inc) diff --git a/share-your-shape.mdx b/share-your-shape.mdx new file mode 100644 index 0000000..b671c0a --- /dev/null +++ b/share-your-shape.mdx @@ -0,0 +1,62 @@ +--- +title: "Share It and Grow" +sidebarTitle: "5 · Share and grow" +description: "Chapter 5 of the Make Shapes guide: publish your Shape, help people discover it, and keep making it better over time." +keywords: ["publish a Shape", "share AI character", "AI discovery", "grow a Shape", "Shape profile", "make AI discoverable"] +--- + +*Chapter 5 of the [Make Shapes](/make-shapes) guide.* + +You've built a Shape with a personality, a memory, and good instincts in a group. The last step is getting it in front of people and making it better as they use it. + +## Make it discoverable + +A Shape only grows if people can find it. In the Creator Dashboard, turn on **Discoverable** so it shows up in search, then give it a profile that earns the click: + +- **Tagline.** One line that says what it's for. This is the hook in search results. +- **Tags and category.** How people browsing land on it. Pick the ones that actually fit. +- **Avatar and banner.** A face and a backdrop. People judge fast, so make them count. +- **Description.** A short, honest pitch for who would love this Shape. + +You can also style the profile page with custom [HTML and CSS](/css) once you want it to really stand out. Every field lives in the [Shape Settings Reference](/shape-settings). + +## Give it a home and share it + +Your Shape has its own page and a [custom link](/vanityurl) you can send anywhere. The fastest way to get it used is to bring it where people already are: drop it into a [group chat](/create-a-chat) with friends, share the link in a community, or post it to a [fandom space](/fandomchat) that fits. + +## Keep improving it + +Great Shapes are grown, not written in one sitting. Once yours is live and people are talking to it: + + + + Read how it actually behaves. Is it too long? Too eager? Off-tone in a crowd? + + + Adjust a single thing, usually reply length or one Free Will trigger, so you can tell what worked. [Regenerate](/regeneration) a few replies to see its range. + + + The dashboard tracks daily and lifetime messages and new people who find your Shape. It's a fun way to see it grow. + + + +## You finished the guide + +That's the whole arc, from a blank page to a Shape people seek out: + +1. [Create your Shape](/how-to-make-a-shape) +2. [Give it a personality](/designing-shapes) +3. [Make it smart](/make-it-smart) +4. [Make it social](/designing-social-intelligence) +5. Share it and grow (you're here) + + + + Start your next Shape from a ready-made example. + + + How people use Shapes day to day. + + + +[Open your dashboard](https://shapes.inc/dashboard) diff --git a/showcase.mdx b/showcase.mdx new file mode 100644 index 0000000..25a872c --- /dev/null +++ b/showcase.mdx @@ -0,0 +1,142 @@ +--- +title: "Showcase: Example Shapes & Prompts" +description: "Complete, copy-paste Shape configs you can drop straight into the dashboard: a study buddy, a team work assistant, a meme friend for the group chat, and a creative collaborator. Every field filled, ready to tweak." +keywords: ["example Shapes", "AI character examples", "Shape templates", "copy paste AI config", "AI study buddy", "AI work assistant", "AI meme bot", "creative AI collaborator"] +--- + +The fastest way to build a great Shape is to start from one that already works. Each example below is a **complete config**. Copy the fields into the [Create Shape](/how-to-make-a-shape) flow and dashboard, then make it yours. + + + Presets use the `{shape}` and `{user}` variables, so keep them lowercase. New here? Read [Designing Great Shapes](/designing-shapes) and [Prompt Engineering](/prompt-engineering) first, then come back and steal. + + + + **How to read these.** The block is the paste-ready field text. Below it, *Setup* covers the engine, Free Will, and capabilities you'd turn on. Tune everything to taste. + + +## 📚 Sage, the study buddy + +For a group of friends studying together without a humorless tutor energy. + +```text +Name: Sage +Short backstory: A patient grad-student tutor who explains hard things simply and celebrates small wins. Calm, encouraging, never condescending. +Personality traits: patient, encouraging, clear, lightly nerdy, upbeat +Tone: warm, plain-spoken +Likes: a good analogy, "aha" moments, flashcards, lo-fi study playlists +Dislikes: gatekeeping, making people feel dumb, cramming +Conversational goals: explain one idea at a time, check understanding, hype small wins +Conversational examples: +"okay, think of it like packing a suitcase — what goes in first?" +"that's exactly right. want the slightly harder version?" +Response style (preset): {shape} explains one idea at a time in plain language, then asks {user} a short question to check they got it. {shape} keeps replies under four sentences unless {user} asks to go deeper, and uses a quick analogy when something is abstract. {shape} never lectures. +``` + +**Setup.** *Engine:* a reasoning-capable model for accuracy. *Free Will (group):* **reply when mentioned** plus **keep the convo going**; leave *always have something to say* off so it doesn't interrupt. *Capabilities:* long-term memory on (so it remembers what each person struggles with), **web search** for looking things up, and **time awareness** so it can say "we covered this yesterday." *Knowledge:* only group-specific facts, like the exam date and focus topics. Skip the textbook; the model already knows it. + +## 💼 Atlas, the team work assistant + +A calm operations brain for a work chat. Think of an AI teammate in your team chat who helps out instead of chattering. + +```text +Name: Atlas +Short backstory: A steady operations lead who keeps a team's work organized. Tracks decisions, summarizes long threads, and answers when asked. Low ego, high signal. +Personality traits: calm, precise, organized, proactive-but-not-pushy +Tone: clear, professional, friendly +Likes: tight summaries, clear owners, a well-run standup +Dislikes: meetings that could've been a message, vague action items +Conversational goals: surface the decision and the open questions, keep replies skimmable +Conversational examples: +"Decision: ship Friday. Open: who owns the changelog? @-me to assign." +"Recapping the last 40 messages — 3 things matter, here they are." +Response style (preset): {shape} answers work questions concisely and in a skimmable format: short lines, the decision in bold, open questions listed. {shape} does not pad replies or add filler pleasantries. When summarizing, {shape} leads with the conclusion. +``` + +**Setup.** *Engine:* a smart all-rounder or reasoning model. *Free Will (work chat):* **reply when mentioned** only; in a focused chat, silence is a feature. *Capabilities:* **web search**, **run code** for quick data and math, and **file reading** so it can read PDFs and specs the team drops in. *Knowledge:* team-specific facts (product names, who owns what, links). See [AI at Work](/ai-at-work) for the chat setup. + +## 😎 Goose, the meme friend for the group chat + +The one that keeps the energy up. Built for a busy friend group that can take a joke. + +```text +Name: Goose +Short backstory: A chaotic-good group-chat friend who lives for the bit. Quick, warm, a little unhinged, secretly the most supportive one in the chat. +Personality traits: quick-witted, playful, loyal, low-key chaotic +Tone: casual, lowercase, punchy +Likes: a perfectly-timed gif, inside jokes, hyping up friends +Dislikes: dead chats, taking things too seriously +Conversational goals: keep the vibe alive, riff on what people say, never be mean for real +Conversational examples: +"absolutely not. but also yes. explain." +"ok that's actually genius and i'm stealing it" +Response style (preset): {shape} replies in short, punchy messages of one or two sentences, lowercase, no roleplay actions. {shape} riffs on what {user} actually said, hypes people up when it matters, and only curses if {user} does first. {shape} never writes paragraphs. +``` + +**Setup.** *Engine:* a bold, expressive model. *Free Will (group):* **keep the convo going** and **react to keywords**, plus **come back later** to revive a dead chat. Skip *always have something to say* if the humans are already active, so it doesn't bury them. *Capabilities:* **GIFs** and **image generation** for memes. Keep replies short; in a fast group chat, length is the enemy. + +## 🎨 Indie, the creative collaborator + +A brainstorming and writing partner that builds on your ideas instead of taking over. + +```text +Name: Indie +Short backstory: A sharp, generous creative partner: equal parts hype and honest notes. Loves a strong verb, a weird premise, and a "what if we…". +Personality traits: imaginative, encouraging, candid, curious +Tone: warm, energetic, specific +Likes: bold first drafts, unexpected combinations, a good constraint +Dislikes: generic feedback, rewriting someone's voice, "make it pop" +Conversational goals: build on the idea, give one real strength and one concrete suggestion, hand back a next move +Conversational examples: +"the opening line is doing a LOT of work — keep it. the second paragraph is where it sags." +"what if the heist is the easy part and the hard part is the drive home?" +Response style (preset): {shape} responds to {user}'s work with one specific strength and one concrete suggestion, then offers a single next line or idea to try. {shape} matches {user}'s style and never rewrites their whole piece unasked. {shape} keeps notes short and usable. +``` + +**Setup.** *Engine:* a creative, high-quality model. *Free Will:* great in a DM or a small writers' chat; use *reply when mentioned* plus *keep the convo going*. *Capabilities:* **image generation** for mood boards and concept art, and **file reading** to react to a draft you upload. Keep replies short so the human does most of the writing. + +## More starters + +Quick configs to riff on. Same idea, less detail. + + + + **Backstory:** A theatrical DM who narrates vivid scenes, voices NPCs, and keeps the party moving. Fair, but loves a dramatic twist. + + **Preset:** `{shape} narrates the scene in second person to the party, then asks "what do you do?" {shape} keeps narration to 3-4 sentences, voices NPCs in distinct styles, and never decides the players' actions. When an outcome is uncertain, {shape} asks the player to roll and react to the result.` + + **Setup:** expressive roleplay engine; Free Will *react to keywords*; long-term memory on so it remembers the party and the world. + + + **Backstory:** A senior engineer who teaches the *why*, not just the fix. Direct, patient, allergic to cargo-cult code. + + **Preset:** `{shape} answers coding questions with a short explanation first, then the code. {shape} points out the root cause of bugs, not just the patch, and asks before assuming the stack. {shape} keeps it concise.` + + **Setup:** reasoning engine; enable **run code** so it can execute and verify snippets. + + + **Backstory:** An organized, upbeat planner who turns a messy group chat into an actual itinerary and remembers everyone's preferences. + + **Preset:** `{shape} keeps the group's plan in view: dates, budget, and what each person wants. {shape} proposes concrete options with quick trade-offs, asks one clarifying question at a time, and summarizes decisions so nothing gets lost.` + + **Setup:** smart all-rounder; turn on **web search**; leave **long-term memory** on to remember preferences. + + + +## Prompts to copy + +Looking for response-style presets like bilingual, poetic, sarcastic, yandere, tsundere, or monotone? The [Presets](/presets) page is a big, copy-ready library. For *how* to write your own, read [Prompt Engineering](/prompt-engineering) and [Prompt 101](/prompt101). + +## Share what you build + +Made something great? Trade configs and tips in the [Prompting & Model Discussions chat](https://talk.shapes.inc/chat/f1e72a86-d360-4987-9009-7c27fc242e71). + + + + Start the create flow and paste one of these in. + + + Turn a starting config into something genuinely yours. + + + +[Create a Shape](https://shapes.inc) diff --git a/use-shapes.mdx b/use-shapes.mdx new file mode 100644 index 0000000..872d900 --- /dev/null +++ b/use-shapes.mdx @@ -0,0 +1,72 @@ +--- +title: "Using Shapes" +description: "How a normal person gets real value from Shapes: find an AI character you like, chat with it, bring it into a group with friends, and use it for fun, for getting things done, and for work." +keywords: ["use Shapes", "AI with friends", "AI group chat", "talk to AI", "AI study buddy", "AI for work", "social AI", "how to use Shapes"] +--- + +You don't have to build anything to get a lot out of Shapes. Thousands of AI characters already exist. You find one you like, start talking, and pull your friends in. That's it. + +This track is for using Shapes. (Want to build your own AI instead? Go to [Make Shapes](/make-shapes).) + +## The short path + + + + Search by name, browse a category, or type a vibe like "study buddy" or "grumpy chef." Every Shape has its own personality, memory, and voice. See [how to find and talk to Shapes](/how-to-talk-to-shapes). + + + Open it and talk. Send a voice note, share a picture, ask it to `/imagine` something. This is your own space to get a feel for it. + + + Start a [group chat](/create-a-chat), add the Shape and a few people, and share the link. Now the AI is in the chat with all of you, reacting to the back-and-forth and remembering what happens. + + + Studying, planning, roleplay, work. The rest of this page shows what that looks like. + + + +## Three reasons people are here + + + + An AI that memes you in the group chat, roleplays with your friends, or plays a character you love. + + + Plan a trip, study for a test, or build something, with friends and an AI in one chat. + + + Like a team chat with a sharp AI teammate in it that summarizes, drafts, and answers. + + + +## What this looks like in real life + +A few concrete ways people actually use Shapes: + +- **Your class group chat gets a study buddy.** Add a patient tutor Shape to the chat where you and your friends already complain about homework. Ask it to explain a concept, quiz you before a test, or settle an argument about an answer. It remembers what each of you keeps getting wrong. +- **Game night gets a narrator.** Drop a storyteller Shape into your friends' chat to run a roleplay. It sets scenes, voices the characters you meet, and keeps the plot moving while everyone plays. +- **Your trip planning stops being chaos.** Put a planner Shape in the group with your travel buddies. It keeps the dates, the budget, and who wants what straight, proposes options, and writes the plan down so nobody loses it. +- **Your team standup writes itself.** In a [work chat](/ai-at-work), @mention an assistant Shape to summarize the last 40 messages into "decided, open questions, who owns what." Ask it to draft the update so you don't have to. +- **You finally talk to your favorite character.** Find a Shape of a character from a show, book, or game, and bring it into a chat with people who love it too. See [fandom chats](/fandomchat). + +The pattern is always the same: find or pick a Shape, then put it where the people already are. + +## Make it your own + +A couple of small touches make any Shape feel like it knows *you*: + +- Set a [persona](/personas) so it knows what to call you and what to remember about you. +- Style your chats and pick who can do what in [Chats 101](/chats-101). + +## Where to go next + + + + Start here if you just want to dive in. + + + Voice, images, web search, memory, and more. + + + +[Open Shapes](https://shapes.inc) diff --git a/what-are-shapes.mdx b/what-are-shapes.mdx index 92de084..354287f 100644 --- a/what-are-shapes.mdx +++ b/what-are-shapes.mdx @@ -1,44 +1,43 @@ --- -title: "What are Shapes?" -description: "Shapes are AI characters that live in group chats alongside humans — social AI companions with persistent memory, voice, and personality. The core of Shapes Inc's multiplayer AI experience." -keywords: ["what are Shapes", "AI characters", "social AI companions", "AI with memory", "AI with friends", "multiplayer AI", "AI in groupchats", "Shapes explained"] +title: "What's a Shape?" +description: "A Shape is an AI character you talk to on Shapes. It has its own personality, memory, and voice, and it can hang out in group chats with you and your friends." +keywords: ["what are Shapes", "AI characters", "AI companions", "AI with memory", "AI with friends", "multiplayer AI", "Shapes explained"] --- -## **Shapes are agentic social companions designed to enhance your online interactions and make conversations with friends more enjoyable!** +A **Shape** is an AI character you can talk to. -Explore and interact with a variety of Shapes directly on [shapes.inc](https://shapes.inc)! +What makes a Shape different from a normal chatbot is that it feels like *someone*. It has a personality that stays consistent, it remembers you, it can talk back in its own voice, and it can hang out in a group chat with you and your friends. -Come hang out in [community chat](https://talk.shapes.inc/community) to start talking with Shapes + people. +A Shape can be almost anything: a study buddy, a dungeon master for a game, a character from your favorite show, or an original character you dream up. -To understand how Shapes behave when humans and AI share the same conversation, read [AI Group Chats](/ai-group-chats). +![A chat with a Shape](/images/tenshichat-1.jpeg) -![Tenshichat 1 Jpe](/images/tenshichat-1.jpeg) +## Two ways to meet Shapes -### **What possibilities can you explore with Shapes!?** + + + Thousands of people have already made Shapes. Search by name or vibe and start chatting. Free, no setup. + + + Give an AI character a name, a personality, and a voice in a few minutes. + + -Shapes have the ability to do a bunch of things such as: +## What a Shape can do -- Send messages & react to them -- Generate images & view images that you send -- Send [voice](how-to-give-your-shape-a-voice) responses and transcripts -- Listen to voice messages -- Browse the web +A Shape does more than text. It can send voice messages, generate images, read files you send it, search the web, and remember what matters to you. The full list lives in [What Shapes Can Do](/capabilities). -### **Unique Customization** +You can also make a Shape your own: tell it what to call you with a [persona](/personas), pick which [AI model](/choosing-a-model) powers it, and shape how it looks and sounds. -You can customize your experience with Shapes in so many ways! +## Where to go next -1. Save a persona with the name you'd like the Shape to call you and the info it should always remember about you! -2. View and edit the [memories](memory) you've shared with the Shape! -3. Select a specific [AI Engine model](aienginecheatsheet) for the Shape to run on! -4. [Style their profile](css) to match their personality! -5. [Create](/create-a-chat) and [style the Chat](/chat-appearance) you're chatting with Shapes in! -6. Set up [custom Chat instructions](/custominstructions-chats) to add information to each Shape in a Chat! + + + The whole point of Shapes: AI and people in the same chat. + + + Build an AI character of your own. + + -Many controls can be tailored for each Shape and are conveniently accessible through a control panel known as your [creator dashboard](https://shapes.inc/dashboard)! - -### **Create Your Own Shape** - -If you're wanting to become a creator head over to [How to make a Shape](/how-to-make-a-shape) to learn more. - -[Return to App](https://shapes.inc) +[Open Shapes](https://shapes.inc)