Skip to content

[INFRA] No Load Balancer or Worker Pool for Concurrent Requests #175

@shrixtacy

Description

@shrixtacy

Description

No request queue, load balancer, or worker pool. Each FastAPI request runs the full pipeline in-process. Under concurrent load, model API calls exhaust connections, rate limits stack up, and memory grows with in-memory caches. mq_agent.py worker exists but is not integrated into the web backend.

Component

System Architecture, ai_council/worker/mq_agent.py

Acceptance Criteria

Integrate load balancing and the existing message queue worker for handling concurrent async queries. Distribute model API calls across worker pools.

Metadata

Metadata

Assignees

Labels

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions