namangoyal88/multi-agent-team

This local-first research application turns a question into a planned, evidence-backed research package. A Streamlit user submits and monitors a job through the API; the workflow asks the NVIDIA-backed Director to plan optional stages, retrieves literature and invokes Groq specialists, optionally runs a restricted experiment, then performs mandatory verification, reporting, and evaluation with optional evidence refinement. SQLite-backed workspace and memory retain job data and prior research. The README documents the major stages and integrations; some wiring is inferred from that description because workflow and API excerpts are partial.