Fix Critical AI Agent Bug

An urgent connectivity fault in the backend/API has brought my AI-driven document-management agent to a stand-still. The rest of the stack—RAG pipeline, vector store, and agent logic—remains untouched, and there have been no recent API changes, so the problem appears to be hidden somewhere between the Python service layer and the LLM endpoint handshake. I need an engineer who has already wrestled with production-grade AI agents, LLM APIs, and retrieval workflows to jump in, isolate the root cause, and restore stable service. You will have direct access to the repo, logs, and monitoring dashboards; I can pair for context if that speeds things up. A lightweight but thorough code review of the surrounding modules is expected so that the final patch doesn’t introduce regressions. Deliverables • Diagnosis summary pinpointing the exact failure path • Clean, tested fix committed via PR (Python) • Confirmation that existing RAG queries and database calls remain unaffected • Brief post-mortem with preventive recommendations Please reply with one or two similar debugging projects you’ve handled, your earliest availability (today is ideal), and whether you prefer an hourly or fixed-price arrangement. I’m ready to give you the green light as soon as we align on scope and timeline.