Summary

FinSight AI is an open-source AI equity research agent focused on building dependable AI research systems. It addresses the infrastructure gap in typical RAG demos by providing resilient workflow orchestration, explicit state transitions, and evidence traceability. Targeting engineers and researchers, it transforms financial data into source-grounded answers and versioned AI reports, offering an auditable approach to AI-driven analysis.

How It Works

This project emphasizes backend infrastructure for AI agents, showcasing resilient workflow orchestration with explicit state transitions for tasks like data ingestion, indexing, and report generation. It employs robust concurrency controls, including Redis Lua single-flight leases and idempotency keys, ensuring task reliability. PostgreSQL with pgvector enables hybrid retrieval (keyword/vector search) and maintains a traceable evidence layer. The system integrates a Spring Boot backend with a FastAPI AI service (optionally using Ollama), supporting versioned reports and RAG evaluation.

Quick Start & Requirements

The full stack is launched via Docker Compose.

• Prerequisites: Docker Engine with Docker Compose v2; 8GB+ RAM recommended for the full stack.
• Installation: Clone the repo and run docker compose up -d --build.
• Optional: Ollama for local LLM, Java 17/Maven 3.9+ (backend dev), Python 3.12 (AI service dev).
• Demo: Execute ./scripts/quick-demo.sh.
• Access: Dashboard/API at http://localhost:8080.

Highlighted Details

• Agent Workflow: Implements recoverable stages for data ingestion, indexing, intelligence build, and AI report generation.
• Concurrency Control: Features idempotency keys, Redis Lua single-flight leases, fencing tokens, and local fallback locks.
• AI Cache: Report caching is tied to data snapshot hashes, preventing reuse of stale conclusions.
• Hybrid Retrieval: Leverages PostgreSQL JSONB, full-text search, and pgvector for merged keyword and vector search.
• Evaluation: Includes RAG and agent quality evaluation for regression checks (hit rate, coverage, hallucination risk).

Maintenance & Community

No specific details on maintainers, community channels (e.g., Discord, Slack), or roadmap were found in the provided README snippet.

Licensing & Compatibility

The license type and compatibility notes for commercial use are not explicitly stated in the provided README snippet.

Limitations & Caveats

The AI service defaults to deterministic, rule-based analysis if Ollama is unavailable, ensuring demo functionality. The project prioritizes backend infrastructure for AI agents, with the UI serving as a demonstration surface.