AI · AI Workflow / Software Engineering module project
Multi-LLM Context Orchestrator
Move active work between AI tools without losing context.
Problem
Switching between LLM tools (ChatGPT, Claude, Gemini…) means re-explaining everything. The history, intent, and constraints get lost, so you waste time rebuilding context in each new tool.
Approach
I built an MVP with a Chrome extension (Manifest V3) that captures an ongoing conversation, and a FastAPI backend running an extract → compress → translate pipeline: it pulls out the key intent and constraints, compresses the context, and rewrites it into a structured handoff prompt tailored to the target model. SQLite stores sessions so handoffs are resumable.
Outcome
A working tool that reduces context loss when switching AI tools, producing a ready-to-paste handoff prompt for ChatGPT, Claude, Gemini, Codex, and generic LLMs.
Tech stack
A browser extension (MV3, JS/TS) handles capture; a FastAPI/Python service runs the extract–compress–translate workflow; SQLite provides resumable session storage. Containerized with Docker and deployed on Render.