KonfAI Studio

A single chatbot (à la ChatGPT / Claude, specialized for medical imaging) that drives konfai-mcp end to end. A clinician-researcher points it at their own dataset and — from the conversation alone — onboards data, authors/reuses a model, trains, infers, visualizes results, compares, keeps & reproduces experiments, then deploys the frozen model privately (on-prem or 100% in the browser). The compute stays on the user's site; nothing is uploaded to a third party.

This is a product surface, not a new engine. Every capability maps 1:1 onto an existing konfai-mcp tool (56 today). The build is the web UI + a thin bridge (BFF), plus the ONNX export.

Layout

• konfai_studio/ — the Python package (the BFF)
  • server.py — FastAPI: streams the chat over SSE, serves the front, streams volumes to NiiVue
  • agent.py — the pluggable brain (KONFAI_STUDIO_LLM): claude-code (Claude Agent SDK, default), openai (local vLLM/Ollama or any OpenAI-compatible endpoint), anthropic (Claude API)
  • web/ — the built front (index.html + assets/, git-ignored; logos are committed)
• frontend/ — the React + Vite source (chat panel + NiiVue viewer; npm run build emits into web/)
• docs/ — the spec, the build plan, and the strategy briefs

Run

pip install -e ./studio                 # deps: fastapi, uvicorn, fastmcp, claude-agent-sdk
npm --prefix studio/frontend install    # once
npm --prefix studio/frontend run build  # builds the front into konfai_studio/web/
konfai-studio                           # -> http://127.0.0.1:8730

The default brain uses your Claude Code subscription (no API key). For a local model: KONFAI_STUDIO_LLM=openai KONFAI_STUDIO_LLM_BASE_URL=http://localhost:11434/v1 KONFAI_STUDIO_MODEL=qwen2.5:14b konfai-studio. Front hot-reload during development: npm --prefix studio/frontend run dev (proxies to the BFF).

See docs/STUDIO_SPEC.md and docs/PLAN.md.