FinSight

Your intelligent companion for financial research — precise insights from SEC filings at your fingertips

Comment

FinSight

Your intelligent companion for financial research — precise insights from SEC filings at your fingertips

FinSight is an AI-powered financial research agent that helps users explore public company performance using official SEC filings. Powered by a multi-agent system, it translates complex questions into structured sub-queries and delivers fact-based insights using real-time data from the SEC EDGAR XBRL API.

🚀 Key Features

  • 🔍 Natural Language Financial Research
    Ask high-level questions like:
    “Are there signs of financial distress in Nvidia?” or
    “Is Apple increasing its investment in R&D?”

  • 🧠 Multi-Agent Architecture

    • Main Research Agent understands intent, decomposes questions, and orchestrates sub-tasks.
    • SEC Filings Expert agent handles structured queries using SEC’s XBRL database.

  • 🧰 Built-In Tools

    • get_company_cik_number: Lookup CIK from company name.
    • get_company_facts_information: List relevant financial facts available for a company based on keyword search.
    • get_available_frames_in_fiscial_year: Retrieve reported frames for a fiscial year.
    • get_data_for_frame: Retrieve data for a specific frame.

  • 🔗 Data Source
    Live integration with the SEC’s public XBRL company concept API.

  • 📖 Explainable Results
    Each response includes specific fiscal year, form type (10-K/10-Q), and filing dates.

Architecture

The system consists of:

  • SecAgent (A2A Service): A worer agent for information retrieval from US SEC.gov. This agent is powered by the SEC EDGAR MCP Server Tools .
  • HostAgent (A2A Service): A coordinator agent that breaks the complex user questions into information retreiving questions.
  • Streamlit UIs:
    • Embedded Runner UI: The UI runs the HostAgent in the same process.

💼 Example Questions You Can Ask

  • “What is Nvidia’s free cash flow trend?”
  • “Does Apple show signs of financial distress?”
  • “How has Tesla’s R&D investment changed in the last 5 years?”

- “What is the most recent earnings per share reported by Amazon?”

Try this project at

https://finsight-n6madsvterjauzntjxefth.streamlit.app/

Running the Full System

This is the recommended way to run the entire multi-agent system.

  1. Start all agent services: A convenience script is provided to launch the Sec agent in the background. The logs will be saved to the logs/ directory.

    python scripts/start_agents.py

    You can check the status of the agents with ps aux | grep _agent.

  2. Run a UI Application:

- **For the Embedded Runner UI:**
  ```bash
  streamlit run ui/app.py --server.port 8080
  ```
  1. Stopping the services: You can stop the background agent processes with: bash pkill -f "_agent"

Testing Individual Agents (Standalone)

For development and debugging, it's crucial to test each agent independently. Each agent has a test_client.py that communicates directly with its A2A service.

  1. Start the target agent's service. For example, to test the Notion agent:

    python -m sec_agent --port 8002

  2. In a separate terminal, run its test client:

    python sec_agent/test_client.py

📦 Tech Stack

  • 🧩 Google ADK
  • 🧠 Gemini LLMs
  • 📄 SEC XBRL API (data.sec.gov)
  • 🔌 Multi-agent orchestration
  • 📜 Python async, logging, environment-driven config

📈 Future Improvements

  • Add support for multi-company comparisons
  • Integrate with vector database for long-form filing summarization
  • Include industry benchmarks for richer insights
  • Stream responses in real-time via web UI or chat interface

🛡️ Disclaimer

FinSight retrieves only public data from the SEC and does not provide investment advice or trading recommendations.

🧑‍💻 Author

Sai Rohit Uddagiri
Data Engineer & AI Agent Developer
LinkedIn | GitHub

Built With

Share this project:

Updates