Financial Agent UI
https://github.com/virattt/financial-agent-ui- Category
- Finance
- Rank
- No. 1562Tools index
Previous survey · No. 1571 ·
- Pricing
- Open Source
- Type
- APP
- Builder
- virattt
- GitHub
- 795 stars
- Date
About
Generative UI for a financial analysis agent. Streams charts, tables, and commentary inline as the agent works through a query.
What it does
A self-hosted research application turns finance questions into a model-directed workflow. The model can fetch prices, financial statement line items, web results, or insider transactions, then return analysis through a browser-based chat interface.
Why it's ranked here
Worth studying as a compact example of model-driven financial research with a custom interface. Its four focused tools, explicit workflow graph, and separate frontend and backend make the architecture easy to follow. The narrow execution path and service requirements keep it closer to an exploration project than a complete research platform.
What's good
The tool selection is concrete and useful: market prices, detailed financials, web search, and insider activity. The model runs deterministically, and the workflow distinguishes direct answers from tool-backed responses. Docker provides a documented route for starting both application halves together.
Tradeoffs
Running it requires credentials for three external services. The workflow executes only the first requested tool and finishes after that tool step, so the shown backend does not perform a second model pass over retrieved results. The page also enforces a 600-pixel minimum content width, which limits narrow-screen layouts.
How to use it well
Use it as a reference implementation or local prototype for engineers exploring generative interfaces around equity research. Docker is the preferred setup, while separate frontend and backend instructions support manual development. Treat its output as exploratory material, not financial advice or a substitute for independent research and professional guidance.
Technical notes+
The Python workflow in backend/gen_ui_backend/chain.py builds a LangGraph StateGraph, binds ChatOpenAI(model="gpt-4o", temperature=0, streaming=True) to four LangChain tools, parses tool calls with JsonOutputToolsParser, and invokes only the first parsed call before ending. frontend/utils/server.tsx converts LangChain streamEvents into React Server Component streams with createStreamableUI and createStreamableValue; frontend/utils/client.tsx exposes serialized actions through React context. frontend/app/page.tsx mounts the chat UI, while frontend/next.config.mjs publishes the backend URL and enables polling for watched files. backend/hot_reload.py restarts the Python server after Python source changes.
Observed
- License
- MIT License
- Languages
- Python backend and TypeScript/TSX frontend
- Install surface
- Repository clone with Docker Compose recommended; manual frontend and backend setup is also documented
- Interface
- Self-hosted browser chat application with separate frontend and backend
- External services
- Requires OpenAI, Financial Datasets, and Tavily API credentials
- Frontend stack
- Next.js, React Server Components, Tailwind CSS, and MUI X Charts
- Backend stack
- LangChain, LangGraph, and LangChain OpenAI
Read from README.md, backend/hot_reload.py, frontend/next.config.mjs, frontend/postcss.config.mjs, frontend/tailwind.config.ts, frontend/lib/mui.ts, frontend/app/page.tsx, frontend/lib/utils.ts, frontend/app/layout.tsx, frontend/app/shared.tsx, frontend/utils/server.tsx, frontend/styles/colors.ts, frontend/utils/client.tsx, backend/gen_ui_backend/chain.py.
What it can do
Stream financial charts in real-time
Financial data query → Live-updating charts and graphs
Generate financial data tables
Financial analysis request → Structured data tables
Provide live financial commentary
Financial query or analysis topic → Real-time text commentary and insights
Process financial analysis queries
Natural language financial questions → Comprehensive financial analysis
Stream multi-format financial reports
Financial analysis request → Combined charts, tables, and commentary in real-time
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.