CuriouslyCory's Developer/Personal Website

• Step 1: Clone the repository
• Step 2: Run npm install
• Step 3: ????
• Step 4: Profit

Blog Feature

The website includes a custom blog system with the following features:

• Custom Post Pages: Each blog post is a custom Next.js page component with its own unique layout and features
• Database Indexing: Custom posts are still indexed in the database for search and listings
• Automated Indexing: Blog indexing is decoupled from the build and runs automatically via the blog-indexing GitHub Action on merge to main. pnpm blog:index still works for local/dev use to extract metadata and searchable body content from custom post components and update the database. If OPENAI_API_KEY is set, the same run also (re-)embeds each post for semantic search — indexing is skipped for posts whose content hasn't changed since the last run, so a normal re-run costs no API calls; set FORCE_REEMBED=1 to force every post to re-embed regardless (useful after changing embedding logic). Editing a post's body requires re-running the indexer for the change to be searchable either way.
• Rich Features: Full support for pagination, tags, search, featured posts, and more
• SEO Friendly: Automatic metadata generation for each post

GitHub Action & Repo Secrets

Indexing runs in the Blog Indexing GitHub Action (.github/workflows/index-blog.yml): it triggers on push to main when files under src/app/blog/**, src/scripts/**, or prisma/** change, and can also be run manually via workflow_dispatch with a force input that sets FORCE_REEMBED to re-embed every post. The workflow applies pending Prisma migrations (pnpm db:migrate) and then runs pnpm blog:index.

The workflow needs the following repo secrets, set under Settings → Secrets and variables → Actions:

Secret	Required	Purpose
DATABASE_URL	Yes	Connection string used by db:migrate and blog:index
OPENAI_API_KEY	No	Enables semantic-search embedding during indexing; without it, indexing is FTS-only
DEFAULT_AUTHOR_ID	No	Fallback author ID for posts that don't specify one

OPENAI_API_KEY must also be set in the Vercel project's environment variables — the GitHub secret only covers indexing, while query-time semantic search on /blog?q= runs in the deployed app.

Search

/blog?q= searches post titles, excerpts, and body text using Postgres full-text search, and — when OPENAI_API_KEY is configured — blends in semantic (meaning-based) search so conceptual queries can find a relevant post even without matching keywords. The two result sets are merged into a single ranking; if no API key is set, or a query is too short, or the semantic lookup fails or times out, search transparently falls back to full-text search alone. Only published posts are matched, and only content as of the last time pnpm blog:index was run — if you edit a post's body, re-run the indexer for the change to show up in search results.

How to Create a Blog Post

1. Create a new directory under src/app/blog/your-post-slug/
2. Create a page.tsx file in that directory with your custom content
3. Include a metadata block as a plain block comment near the top of the file. The opening marker must be ---bm (this is what the indexer's parser looks for), and published: true is required — without it the post is treated as a draft and never appears in listings or search:

/*
---bm
title: Your Post Title
excerpt: Short description of your post
coverImage: /images/blog/your-image.jpg
publishedAt: 2023-12-15
featured: true
published: true
tags: tag1,tag2,tag3
--- 
*/

The slug is taken from the directory name, not the metadata block. The block is parsed as YAML, so wrap any value containing a colon in quotes (e.g. title: "My Post: A Subtitle") — an unquoted : is read as a key/value separator and breaks parsing.

4. Merge to main and the blog-indexing GitHub Action will automatically index the new post (search and listings won't show it until then). The indexer extracts both metadata and searchable body content from the page component, so a merge that edits a post's body triggers re-indexing too. Before merging, run pnpm blog:index locally to verify the post extracts cleanly. Blog indexing is decoupled from the build. With OPENAI_API_KEY set, the same command also embeds the post for semantic search — the cost for a full blog's worth of posts is a few pennies at most, and unchanged posts are skipped automatically on re-runs (set FORCE_REEMBED=1 to bypass that and re-embed everything).

Development Commands

Command	Description
pnpm dev	Start the local dev server with Turbopack
pnpm lint	Run ESLint
pnpm typecheck	Run the TypeScript compiler in --noEmit mode
pnpm test	Run the Vitest test suite
pnpm build	Build for production (does not touch the database)
pnpm db:migrate	Apply Prisma migrations (prisma migrate deploy) — source of truth for schema changes; prefer this over prisma db push
pnpm blog:index	Run the indexer: extract metadata and searchable body content from custom post components and update the database. Also (re-)embeds posts for semantic search when OPENAI_API_KEY is set, skipping any post whose content is unchanged since its last embed

Credits

This is a T3 Stack project bootstrapped with create-t3-app.