FinancialAgentia is an evidence-led financial research application with a React browser workspace, FastAPI API, Redis-backed research jobs and an independent Python worker. It also includes financial-data inspection, explicit saved research context, read-only portfolio analysis, periodic monitoring and permissioned team research.
Current supported runtime: Bun-built web assets served by FastAPI → Redis durable execution → Python research worker → configured AI and financial providers. The older Bun/Ink terminal client remains available as a backend client; it is not a second research engine.
- Documentation index
- Quick start: Docker or manual
- Windows PowerShell setup
- Linux/macOS setup
- Configuration, authentication and providers
- Browser feature user guide
- Operations, upgrade and backup
- Troubleshooting
- Production deployment
- Tests and verification status
Requires Docker with Compose, OpenAI or a reachable Ollama model, and licensed Financial Datasets credentials for market research. From the repository root:
git clone https://github.com/SteveParadox/FinancialAgentia.git
cd FinancialAgentia
Copy-Item python-backend\.env.example python-backend\.env
# Edit python-backend\.env: set OPENAI_API_KEY, LLM_MODEL, FINANCIAL_DATASETS_API_KEY.
docker compose -f docker-compose.research.yml -f docker-compose.local.yml config
docker compose -f docker-compose.research.yml -f docker-compose.local.yml up --build -d
docker compose -f docker-compose.research.yml -f docker-compose.local.yml psOpen http://127.0.0.1:8000/ and select Open workspace with a blank access token. This local-only overlay deliberately disables authentication, and the API port is bound to the host loopback address. Never deploy the overlay publicly or use it with private production data. Docker's base Compose file is production-oriented, requires properly provisioned authentication, and does not support a blank browser token.
Run docker compose -f docker-compose.research.yml -f docker-compose.local.yml logs -f api worker to diagnose startup. /health checks the HTTP process; /ready requires Redis, a recent worker heartbeat, registered tools and configured model credentials. Neither endpoint proves paid market-data access.
The Windows command Copy-Item above is PowerShell-specific. See the Linux/macOS guide for cp and shell commands.
| Area | Current scope |
|---|---|
| Research | Company, comparison, historical-price and general scoped jobs; progress, cancellation, retry, saved evidence-backed results |
| Model providers | OpenAI and Ollama local/cloud; additional legacy optional adapters subject to compatible packages |
| Financial data | Financial Datasets market/fundamental data and optionally configured SEC filing facts; no blanket claim of live quotations |
| Context | Explicitly saved reports for authorized follow-up; not a general semantic/vector memory service |
| Portfolio | Long-only, single-currency transaction ledger; valuation, scenario shocks and a limited historical risk proxy |
| Monitoring | UTC periodic price crossing rules and recurring research; in-app notifications only |
| Collaboration | Authenticated workspaces, invitations, roles, frozen shared report snapshots, comments, review state and JSON/Markdown/CSV exports |
Important limitations: The app does not implement brokerage trading, self-service web signup, email notification delivery, a general FX conversion engine or verified live-market entitlement. A risk proxy calculated with current quantities is not actual investor performance. Research results and financial data can be partial or unavailable.
The root package.json contains build:web, typecheck:web, test:web, typecheck and test. The Python backend has a separate virtualenv and requirements-dev.txt. Follow the manual setup and verification guide, not the legacy terminal-first installation text.
The companion legacy terminal client is started with bun start once the API and worker are operational; its own credential settings live in the root env.example.
- Research API
- Target architecture
- Existing execution operations
- Existing implementation report
- Current docs validation
Financial research is informational, not a guarantee of investment outcomes.