Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/syhily/yufan.me/llms.txt

Use this file to discover all available pages before exploring further.

yufan.me requires two data stores: Postgres for all content, settings, sessions, and analytics, and Redis for session storage, rate limiting, and generated-image caching. Both must be reachable before the server starts. Schema management is handled by Drizzle Kit — migrations live under drizzle/ and are applied independently of application boot.

​
Postgres

yufan.me targets Postgres 16 or later. Postgres 17 is recommended. The pgvector extension is used for embedding-based search but is optional — the application falls back to SQL LIKE search when the extension is unavailable. Connection strings follow the standard libpq URI format: 
DATABASE_URL=postgres://user:password@host:5432/database
If you are using SSL, append ?sslmode=require (or verify-full for certificate verification): 
DATABASE_URL=postgres://user:password@host:5432/database?sslmode=require
The application connects via the pg driver through Drizzle ORM. All queries go through the pool configured in src/server/infra/db/.

​
Redis

Redis is used for three purposes:
  • Session storage — signed session tokens are persisted in Redis so they survive server restarts.
  • Rate limiting — per-IP and per-user counters enforce the limits configured in blog.rateLimit settings.
  • Caching — generated OG images and other computed artifacts are cached to avoid redundant work.
Connection strings follow the ioredis URI format: 
REDIS_URL=redis://host:6379
For Redis with authentication: 
REDIS_URL=redis://:password@host:6379
Redis Sentinel and Cluster topologies are not required — a single standalone Redis instance is sufficient for most deployments.

​
Running migrations

Migrations are managed with Drizzle Kit. The drizzle/ directory contains one subdirectory per migration, each with a SQL file and snapshot metadata. The directory is copied into the Docker runtime image so migrations can be applied from within a container without access to the source tree.
1

Set DATABASE_URL

Ensure DATABASE_URL is exported in your shell or present in your .env file before running any migration commands:
export DATABASE_URL=postgres://user:password@host:5432/database
Drizzle Kit reads this variable automatically. No separate drizzle.config.ts override is needed for the standard deployment path.
2

Generate migrations after schema changes (development only)

When you modify src/server/infra/db/schema.ts, generate a new migration file with:
npm run db:generate
This runs drizzle-kit generate and writes a new timestamped directory under drizzle/. Commit the generated files alongside your schema changes. You do not need to run this step when deploying an existing release.
3

Apply migrations

Apply all pending migrations against your database using drizzle-kit migrate:
npx drizzle-kit migrate
When deploying with Docker, run migrations from a temporary container before starting the application container:
docker run --rm --env-file .env yufan.me \
  npx drizzle-kit migrate
Drizzle Kit tracks applied migrations in a __drizzle_migrations table and skips files that have already been applied. It is safe to run on every deployment.

​
Migration files

The drizzle/ directory contains ten migrations applied in timestamp order:
MigrationDescription
20260514000000_enable_pgvectorEnables the pgvector extension for embedding-based search. Must be the first migration to run.
20260514000001_init_schemaInitial schema: post, page, category, tag, user, setting, image, music, friend, and junction tables.
20260514000002_access_logAnalytics visit ingestion table (access_log) used by the first-party analytics dashboard.
20260514000003_access_log_timescaleConverts access_log to a TimescaleDB hypertable for time-series performance. Falls back gracefully when TimescaleDB is not installed.
20260515153336_audit_logAudit log table that records admin events (content changes, settings updates, user actions).
20260517143442_sidebar_widgetsSidebar widget configuration stored as JSONB under blog.sidebar settings.
20260517150000_x_rebrand_and_footer_navUpdates social link keys to reflect the X (formerly Twitter) rebrand and adds footer navigation columns.
20260517160000_merge_footer_nav_into_navigationConsolidates footer navigation configuration into the blog.navigation settings section.
20260517170000_merge_footer_into_generalMerges footer display settings into the blog.general settings section.
20260521000000_audit_log_indexesAdds performance indexes to audit_log to speed up admin dashboard queries.
The audit_log table is excluded from standard pg_dump backups because it uses S3 archival for long-term retention. If you are running manual backups, add --exclude-table=audit_log to your pg_dump command to avoid backing up data that is already archived to object storage.
The pgvector extension is required for embedding-based search. If your Postgres provider does not support pgvector (for example, a standard cloud Postgres without extensions), the first migration will fail. In that case, comment out or skip 20260514000000_enable_pgvector and the application will automatically fall back to SQL LIKE search across post titles and bodies. All other features remain unaffected.

Build docs developers (and LLMs) love

Get started for freeTalk to us