- scaling.md: Link each step to relevant troubleshooting sections
(DB corruption, WebSocket errors, login loops, worker crashes,
file access issues, logging, OpenTelemetry)
- multi-replica.mdx: Add scaling guide link in intro and Related Docs,
plus links to Redis and RAG troubleshooting
- performance.md: Add scaling guide link in Scaling Infrastructure section
- redis.md: Add scaling guide link in "When is Redis Required?"
- connection-error.mdx: Add links to Redis tutorial, scaling guide,
and multi-replica WebSocket troubleshooting
- Helm.md: Add scaling guide link alongside existing HA guide link
https://claude.ai/code/session_01TPoquFdHG6dZxRrZ4Jormh
- Added Step 4 (Switch to External Vector Database) explaining why default
ChromaDB crashes in multi-process setups, with a comparison table of
alternatives (PGVector, Milvus, Qdrant, Pinecone, ChromaDB HTTP mode)
- Corrected Step 5 (file storage): shared filesystem (NFS) is sufficient,
cloud storage (S3) is optional. Files use UUID-based names so no write
conflicts occur between processes.
- Updated architecture diagram, env var examples, and quick reference table
to include vector DB column and clarify storage requirements.
https://claude.ai/code/session_01TPoquFdHG6dZxRrZ4Jormh
ChromaDB's default local PersistentClient uses SQLite which is not fork-safe.
When uvicorn forks multiple workers, concurrent writes crash workers instantly.
Added warnings and guidance across env config, HA/scaling, performance,
troubleshooting, Docker Swarm, Helm, Redis, RAG, and enterprise architecture docs.
https://claude.ai/code/session_01TPoquFdHG6dZxRrZ4Jormh
