[GSoC 2026 Proposal] Advanced RAG & Performance Optimization for AI Chatbot — Abhishek Sharma

Hi Jenkins Community and Mentors,

I’m Abhishek Sharma, and I have been actively contributing to the resources-ai-chatbot-plugin codebase. My primary focus has been on resolving critical concurrency bottlenecks to ensure the plugin scales effectively under load.

My Contributions So Far: I have successfully merged two PRs addressing event loop starvation:

  • PR #102: Eliminated WebSocket streaming lag by offloading blocking disk I/O using run_in_executor.

  • PR #133: Fixed a major API freeze in the file upload endpoint by wrapping synchronous LLM inference inside asyncio.to_thread.

  • Result: These changes ensured the FastAPI event loop remains responsive to health checks and concurrent sessions during heavy inference.

Proposed GSoC 2026 Direction:

  1. Scalable, Non-Blocking Architecture: Implementing standardized async-safe patterns and a performance benchmarking suite to prevent future event loop regressions.

  2. Advanced Retrieval via Hybrid Graph-RAG: Moving beyond semantic search to a Hybrid RAG model using NetworkX to map Jenkins metadata (plugin dependencies, advisories) for multi-hop reasoning.

  3. Agentic Tooling: Developing an Agentic Tool for automated Jenkins log analysis.

Feedback Request: I would value mentor feedback on whether a lightweight in-memory graph (NetworkX) is preferred over a database like Neo4j for this environment, and which Jenkins APIs should be prioritized for the Agentic Tool layer.

Looking forward to your suggestions!

GitHub: https://github.com/HeyBoY-ops