feat: scaffold browser extension with shared types and formatters#1
Open
ai-diffusion wants to merge 20 commits into
Open
feat: scaffold browser extension with shared types and formatters#1ai-diffusion wants to merge 20 commits into
ai-diffusion wants to merge 20 commits into
Conversation
Adds artifacts/browser-extension/src/lib/ with the foundational types (Message, Conversation, Format) and pure format converters (Markdown, Plain Text, JSON, HTML) for the AI Chat Exporter extension. https://claude.ai/code/session_0138YhQCbBxY5KowePhCK1y3
Adds DOM extractors for ChatGPT, Claude, Grok, Gemini, and Perplexity. Each script listens for EXTRACT_CONVERSATION messages and returns a structured Conversation object from the live page DOM. https://claude.ai/code/session_0138YhQCbBxY5KowePhCK1y3
Dark-themed popup with conversation preview, format selector, and Download/Copy buttons. Handles unsupported pages and empty conversations gracefully. https://claude.ai/code/session_0138YhQCbBxY5KowePhCK1y3
Adds package.json, tsconfig.json, build.mjs (esbuild), manifest.json (MV3, Chrome + Firefox), and a minimal service worker. Build outputs to dist/ via `npm run build`; `npm run pack` produces extension.zip. https://claude.ai/code/session_0138YhQCbBxY5KowePhCK1y3
Covers Chrome Web Store, Firefox AMO, Zen Browser, and sideloading in Chrome/Brave/Edge/Firefox, plus a pre-publish testing checklist. https://claude.ai/code/session_0138YhQCbBxY5KowePhCK1y3
Indigo chat bubble with white download arrow, matching the design from the ChatGPT-generated reference image. All three sizes generated programmatically for consistent rendering. https://claude.ai/code/session_0138YhQCbBxY5KowePhCK1y3
Ready-to-sideload dist/ folder and ready-to-upload extension.zip for Chrome Web Store and Firefox AMO submissions. https://claude.ai/code/session_0138YhQCbBxY5KowePhCK1y3
…tructure - Add htmlToMarkdown() recursive walker: converts h1-h4, bold, italic, code blocks, lists, links to Markdown — preserves formatting instead of flat innerText - Two-tier selector fallback: article[data-testid^="conversation-turn-"] (2024-2025 DOM) then [data-message-author-role] (any build) - Drill into .prose/.markdown child for accurate content extraction, skipping copy/regenerate buttons and SVG icons - Detect pasted attachments (pre, [class*="attachment"], etc.) inside user turns and append them with --- separator - Filter non-user/assistant roles (tool, system) to keep output clean https://claude.ai/code/session_0138YhQCbBxY5KowePhCK1y3
…forms - src/lib/capture.ts: shared engine — autoScrollToTop loads the entire lazy-loaded history before extraction (stability detection, 60-iteration cap, truncated flag, virtualization contingency via window merging); collectAssets registers images/file links; GET_ASSET serves them to the popup (canvas re-encode for blob: images, credentialed fetch fallback) - src/lib/dom-to-markdown.ts: HTML→Markdown walker shared by all platforms (was ChatGPT-only) with onImage hook for inline asset markers - Content scripts reduced to per-platform configs: selector tiers, role detection, prose selectors — Claude/Gemini/Grok/Perplexity now get full markdown structure and full-history capture, not flat innerText - ChatGPT: pasted-text chips probed for hidden full text; honest [Attached: …] placeholder when content isn't in the page - Popup: live "Loading full conversation… (N messages)" progress, 90s watchdog, "Include images & files" checkbox, sequential asset downloads into chat-export-<title>/ subfolder via downloads API - Manifest: cover codex.openai.com https://claude.ai/code/session_0138YhQCbBxY5KowePhCK1y3
Firefox does not support background.service_worker alone in MV3 — adding background.scripts alongside it so the extension loads in both Chrome and Firefox/Zen without error. https://claude.ai/code/session_0138YhQCbBxY5KowePhCK1y3
Updates manifest.json, popup title/header, and package.json name from old placeholder 'AI Chat Exporter' to the agreed name. Also improves manifest description. https://claude.ai/code/session_0138YhQCbBxY5KowePhCK1y3
…Obsidian button - Filename format: YYYY-MM-DD Platform Model Chat-Title.md (e.g. '2026-06-11 ChatGPT GPT-4o My Conversation.md') - YAML frontmatter in all markdown exports: title, date, time, platform, model, url, messages count, tags — Obsidian reads this as note properties automatically - Model detection added to all 5 platforms via modelOf() — scrapes the model selector button, falls back to page title - 'Save to Obsidian' button: opens obsidian://new?file=...&content=... to create the note directly in the running vault; falls back to .md download for very large chats (>50k chars) - Conversation type gains optional model field - Plain text and HTML exports also include model name in metadata https://claude.ai/code/session_0138YhQCbBxY5KowePhCK1y3
- computeStats() calculates totalWords, totalChars, estimatedTokens (chars/4 — ~10-15% accuracy vs real tokenizer), userMessages count, assistantMessages count from conversation content - YAML frontmatter: words, characters, estimated_tokens, user_messages, ai_messages fields added — all searchable/filterable in Obsidian - Markdown body: 'Words: N | Est. tokens: N | Characters: N' line - Plain text: same stats line in header - HTML: stats in the meta bar alongside platform and model - Zip confirmed self-contained: 17 files, no runtime dependencies https://claude.ai/code/session_0138YhQCbBxY5KowePhCK1y3
Filename: YYYY-MM-DD — Platform — Title — exported D Month YYYY, H.MMam [N msgs ~Ntk]
Markdown structure:
- ## File info (source, exported date)
- ## Size (messages, words, chars, estimated tokens)
- ## Cost / context-fit table (8 models, context %, input cost, output rate)
- ## You · N · [preview] / ## Platform · N · [preview] numbered headings
→ Obsidian outlines the full conversation with exchange numbers on right panel
- Exchange counter increments on each user turn so AI·N matches You·N
Platform name used as AI speaker (ChatGPT/Grok/Claude) not generic 'AI'
Friendly dates: 'Thursday, 11 June 2026, 4.39pm'
Cost table prices verified June 2026; note subscription plans = $0/chat
JSON export includes computed stats block
HTML export includes cost table and exchange-numbered message labels
https://claude.ai/code/session_0138YhQCbBxY5KowePhCK1y3
- dom-to-markdown.ts: add full <table>/<tr>/<th>/<td> → markdown pipe table support; separator row inserted after header row (or first row if all <td>); escapes pipes inside cell content - popup.ts: rename assets after extraction to match chat filename format (YYYY-MM-DD — Platform — Title — exported … — image-N.ext); rewrites embedded image refs in message content to use flat filenames; switches text-file download to chrome.downloads.download() so the chat .md and all images land in the same directory https://claude.ai/code/session_0138YhQCbBxY5KowePhCK1y3
capture.ts: add extractFileAssets async hook to PlatformConfig; call it per-turn in captureConversation() after regular asset collection, counting user turns so each file gets a prompt-N label. chatgpt.ts: implement extractFileAssets — finds file/document chips in user turns, tries DOM hidden text first (Tier A), then clicks the chip, waits up to 2.5s for ChatGPT's viewer modal, extracts content from the pre/code element, closes the modal, and registers the text as a Blob URL Asset. Companion files are named "prompt-N — Pasted text(1).txt" which popup.ts prefixes with the chat filename so the final download is "2026-06-11 — ChatGPT — … — prompt-1 — Pasted text(1).txt". Also update the [Attached: …] placeholder to [File: name] (neutral, since the content is now in the downloaded companion file). https://claude.ai/code/session_0138YhQCbBxY5KowePhCK1y3
The extractFileAssets hook added to PlatformConfig is now included in all bundled IIFE content scripts (claude, gemini, grok, perplexity). https://claude.ai/code/session_0138YhQCbBxY5KowePhCK1y3
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adds artifacts/browser-extension/src/lib/ with the foundational
types (Message, Conversation, Format) and pure format converters
(Markdown, Plain Text, JSON, HTML) for the AI Chat Exporter extension.
https://claude.ai/code/session_0138YhQCbBxY5KowePhCK1y3