NeuroDiscover AI by Quorum

Quorum — six agents, one evidence store, ranked commercial discovery. Optional CUA (Conclusion Update Agent) grant-proposal engine with a static NIH demo on dashboard Step 3.

Agentic system for commercial development discovery: continuously ingest literature and trials, surface hidden patient subgroups inside heterogeneous indications, map subgroup → mechanism → treatment connections, and rank opportunities by evidence and commercial potential.

Live docs: neurodiscover.github.io · Track: 02 — Autonomous Research · Partner framing: Pfizer Commercial Development Discovery

Requires Python 3.10, 3.11, or 3.12 on macOS, Linux, or Windows.

pip:

git clone https://github.com/kahinimehta/Quorum.git
cd Quorum/neurodiscover
python3 --version          # must be 3.10.x – 3.12.x
pip install -r requirements.txt
cp .env.example .env
make dashboard    # same as: python3 cli.py dashboard

conda (empty environment):

git clone https://github.com/kahinimehta/Quorum.git
cd Quorum/neurodiscover
conda env create -f environment.yml
conda activate neurodiscover
cp .env.example .env
make dashboard

Opens the URL printed by the launcher (UI on 8080, API default 5000 with auto-fallback; URL always includes ?api= so the static server reaches FastAPI). Press Ctrl+C to stop.

Three tabs: centered column (max 1400px) — Step 1 configure & run · Step 2 discovery results (KPI strip · cohort, rankings, hypotheses; Audit & provenance collapsed) · Step 3 static grant proposal (graded6 — expanded proposal · collapsed Grant pipeline + Pipeline trace & audit; Jump to nav).

Windows: use python cli.py dashboard if make is not installed. SSH / headless: add --no-browser and open the Open: URL from launcher output.

Full setup, platform notes, and flags → docs/dashboard.md · Pipeline modes FAQ · Debugging

Team live database: Supabase holds 300+ real evidence rows. Set SUPABASE_DATABASE_URL in .env and do not run cli.py build (it truncates all tables). Use scan / pull to add evidence only.

Storage: SQLite locally, or Supabase Postgres for the team shared DB (SUPABASE_DATABASE_URL). Table name is evidence, not papers. See docs/developer-reference/database.md and docs/developer-reference/supabase.md.

cd neurodiscover
pip install -r requirements.txt
cp .env.example .env
python3 cli.py build      # local SQLite only — never on team Supabase
python3 cli.py validate
python3 cli.py demo       # safe offline literature-agent run
python3 cli.py pull --disease "your condition" --max 150

Naming: docs and guides use lowercase kebab-case (e.g. docs/developer-reference/agent-io.md). Root README.md, AGENTS.md, and CONTRIBUTING.md stay uppercase — see CONTRIBUTING.md.

Quorum/
  AGENTS.md
  cua/                   # Optional Agent 6 — NIH grant proposal engine (CLI; Step 3 shows static demo)
  docs/                  # Jekyll site (Just the Docs) → neurodiscover.github.io
  queries.sql
  neurodiscover/
    cli.py               # entry point — run all commands from here
    schema.sql, schema.pg.sql, seed_data.json
    db.py, seed.py, paths.py, db_validate.py
    agents/
      literature_agent.py    # Agent 1 (literature synthesis)
    ingestion/
      grants_pull.py         # NIH RePORTER grants
    validation/
      consistency.py, pubtator.py, spot_check.py, report.py
    scripts/
      export_team_keys.py
    models/
      schemas.py             # Person 2 API DTOs (not DB schema)
    api_server.py            # FastAPI dashboard backend (:5000)
    orchestrator.py          # Agents 1→6 pipeline for POST /api/run-discovery
    frontend/
      index.html             # Dashboard UI
      quickstart.md          # Dashboard quick start

Dashboard docs: docs/dashboard.md (UI) · docs/developer-reference/dashboard-api.md (API reference).

Quorum — Alia Merchant · Amy He · Ayelet Peres · Kahini Mehta · William Yakah

Full bios and collaboration map → docs/team.md (docs site)