Blog CQRS service

Build a Node.js (or Go) service that accepts post-creation commands, writes them to PostgreSQL as the source of truth, and maintains a Redis sorted-set projection of the last 20 posts per author. The service must handle concurrent writes from two parallel clients without projection corruption, and serve the feed endpoint exclusively from Redis.

• POST /commands/create-post endpoint: validates DTO, writes to PostgreSQL, updates Redis projection atomically per author
• GET /queries/feed/:authorId endpoint: reads exclusively from Redis, never touches PostgreSQL
• Concurrent write test: script that fires 50 parallel CreatePost commands for the same author and asserts the feed contains exactly 20 items (the most recent) with no duplicates
• Stale-read test: kill Redis mid-run, confirm PostgreSQL writes still succeed, restart Redis, run a rebuild script that re-seeds the projection from PostgreSQL
• README documenting the two-database contract: what PostgreSQL owns, what Redis owns, and what happens when each goes down

• Feed endpoint returns exactly 20 items (not fewer, not more) after 50 inserts for the same author
• Concurrent write test produces zero duplicate postIds in the feed projection across 50 parallel inserts
• POST /commands/create-post returns 422 with a structured error body when title is empty — PostgreSQL and Redis are never touched
• After Redis restart and projection rebuild, GET /queries/feed/:authorId returns the same 20 posts as before the restart
• PostgreSQL INSERT and Redis ZADD happen in a single function with no branching between them — the Redis failure path logs a warning but does not roll back the PostgreSQL write