v0.7.0 — Scalability & Background Jobs

What's New in v0.7.0

🚀 Scales to 100+ Concurrent Users

Async Redis — Switched from blocking redis.Redis to redis.asyncio.Redis. All Redis I/O is now non-blocking, so 100 users don't queue behind a single socket call.

Session Health Checks — Cached MCP sessions are pinged before reuse. Dead subprocesses are auto-removed and recreated on the next request. No more broken-pipe errors.

Multi-Process Serving — Dockerfile now uses gunicorn with 4 uvicorn workers (was 1). Utilizes all CPU cores.

Worker Scaling — docker-compose runs 3 RQ worker replicas (was 1). Background jobs execute in parallel.

Redis Health Checks — docker-compose uses service_healthy condition so BridgeKit waits for Redis to be ready before starting.

🛠 Other Changes

• Request/error counters + get_stats() on /health endpoint
• Memory limits in docker-compose (1G server, 512M per worker)
• Expanded log buffer (100 → 200 entries)
• 8 tests (2 new), all passing
• CI fixed: added uv.lock + [dependency-groups] for uv sync --dev

📦 One-Click Deploy

---

Full Changelog: v0.6.0...v0.7.0