Skip to content

Repository files navigation

Data Research Workbench

An educational public-data research workspace. V1 accepts a question and one or more public sources, runs a FastAPI/Python workflow, and returns provenance, deterministic analysis, confidence, charts, findings, limitations, and JSON/CSV exports.

The Vercel frontend can run Live Mode through a deployed FastAPI backend. If the backend is not configured or unavailable, the app falls back to clearly labeled Demo Mode.

Features

  • Public CSV, REST API JSON, RSS, and GitHub raw connectors
  • Fetch safety: timeout, size limit, row cap, content-type validation, private network blocking
  • Conservative cleaning with logged transformations
  • Deterministic numeric, missing-value, category, date-trend, and correlation analysis
  • IQR outlier detection and source comparison
  • Metadata: source count, row count, missing values, methods, timestamp, confidence
  • Ollama local summarization with template fallback
  • Optional 9Router fallback adapter
  • Temporary V1 run storage in backend memory
  • Export JSON and CSV
  • Architecture page explaining the current workflow and V2 path

Local Setup

Install frontend dependencies:

pnpm install

Install backend dependencies:

python -m pip install -r apps/api/requirements.txt

Run the backend locally:

pnpm api:dev

Run the frontend:

pnpm dev

Open http://localhost:3000.

Environment

Copy .env.example to .env.local for frontend local mode:

FASTAPI_BASE_URL=http://127.0.0.1:8000

Optional local model routing:

OLLAMA_BASE_URL=http://127.0.0.1:11434
OLLAMA_MODEL=llama3.2
NINE_ROUTER_BASE_URL=http://127.0.0.1:20128/v1
NINE_ROUTER_MODEL=auto

Quality Checks

pnpm lint
pnpm typecheck
pnpm test
pnpm build
python -m pytest apps/api/tests -q

Live Deployment

Deploy apps/api as a separate Vercel FastAPI project, then deploy apps/web with:

FASTAPI_BASE_URL=https://data-research-workbench-api.vercel.app

The frontend route /api/research proxies to FastAPI and labels returned runs as Live Mode.

V1 Limits

  • Single-user local mode only.
  • Run history is temporary and may disappear when FastAPI restarts.
  • Frontend falls back to deterministic Demo Mode when no backend is configured.
  • No database, queue, auth, broad scraping, or required OpenClaw integration in V1.

Deployment

The frontend can deploy to Vercel as a Next.js app from apps/web. The production demo remains deterministic unless FASTAPI_BASE_URL points to a deployed backend approved later.

Live Deployment

Demo: Autonomous Data Research Agent on Vercel

About

Semi-autonomous data research agent that collects public data, validates and cleans it, generates analysis and charts, and produces source-backed reports using Python, FastAPI, Ollama, and optional 9Router/OpenClaw integrations.

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages