Laboratree

Grow · Innovate · Impact

The trustworthy, agentic, human-in-the-loop research lab.

---

Laboratree is an end-to-end, multi-agent research lab covering the full lifecycle of primary/secondary data research — from messy raw inputs to consolidated data, to understanding and reproducing research papers, to modelling (ML / DL / econometrics), all with provenance-locked, reproducible results and human-in-the-loop control.

See the foundation plan for the full architecture and rationale.

Monorepo

apps/api          FastAPI backend (uv workspace member)
apps/web          Next.js frontend
packages/plugin-sdk   Component / registry contracts (uv workspace member)
infra             docker-compose + Dockerfiles
data              local BlobStore volume (gitignored)

Tech

• Backend: FastAPI (Python 3.12, managed by uv), LangGraph agent orchestration, Celery.
• Frontend: Next.js / React, React Flow, dnd-kit.
• Persistence (dedicated containers): Postgres (+pgvector), Redis, Neo4j, MongoDB. Blobs on a local volume behind a BlobStore interface.
• LLMs: OpenAI (pluggable LLMClient).

Quick start (dev)

# 1. Bring up the datastores + services
cp .env.example .env            # then fill in OPENAI_API_KEY
docker compose -f infra/docker-compose.yml up -d

# 2. Backend (local, without Docker) — uses uv
cd apps/api
uv sync                          # creates .venv, installs workspace deps
uv run uvicorn laboratree.main:app --reload

# 3. Frontend
cd apps/web
npm install && npm run dev

Health check: http://localhost:8000/health reports connectivity to every datastore.