Skip to content

FinRAG Agent v1.0.0 — Initial Release

Latest
Latest

Choose a tag to compare

View all tags
@zeron-G zeron-G released this 09 Mar 20:59
· 2 commits to main since this release
v1.0.0
23f9776

FinRAG Agent v1.0.0

Course: JHU BU.520.710 AI Essentials for Business — Spring 2026 Final Project (Team 5)

What is FinRAG Agent?

A production-grade Retrieval-Augmented Generation (RAG) system for analyzing SEC 10-K annual filings of Alphabet (Google), Amazon, and Microsoft. Built with a hierarchical dual-chunk pipeline, multi-provider LLM support, and a zero-dependency browser frontend.

✨ Key Features

RAG Pipeline

  • Hierarchical chunking: fine (500 chars) + coarse (2000 chars) with overlap
  • Automatic 10-K section detection (Items 1, 1A, 2, 6, 7, 7A, 8)
  • MMR (Maximal Marginal Relevance) retrieval via FAISS for relevance + diversity balance
  • Fine-to-coarse chunk expansion for context-rich answers

Multi-Provider LLM Support

  • OpenAI: GPT-4o, GPT-4o-mini, GPT-3.5-turbo
  • Google: Gemini 2.0 Flash, 1.5 Flash, 1.5 Pro
  • Anthropic: Claude Sonnet, Claude Haiku
  • Local: Ollama (Llama 3.1, Mistral, Phi-3, any installed model)

Conversation Memory

  • Sliding-window: last 6 turns verbatim + LLM-generated rolling summary of older history
  • JSON-persisted sessions across server restarts

Agent Tools

  • file_read / file_write / file_list
  • pdf_read (PyPDF-based)
  • code_execute (sandboxed Python)
  • data_analysis (pandas/numpy sandbox)
  • web_search (DuckDuckGo)

Frontend

  • Pure HTML/CSS/JS — no build step, no npm
  • Dark mode, session management, source chunk panel, quick-question shortcuts
  • SSE streaming for real-time token-by-token responses

Citation Quality

  • System prompt enforces inline citations: company name, 10-K section, and page number for every financial figure

📁 Included 10-K Documents

Company Filing
Alphabet (Google) 10-K Annual Report 2024
Amazon 10-K Annual Report 2024
Microsoft 10-K Annual Report 2024

🚀 Quick Start

git clone https://github.com/zeron-G/FinRAG-Agent.git
cd FinRAG-Agent
conda env create -f environment.yml
conda activate rag-agent
cp .env.example .env   # fill in your API key(s)
python run.py

Then open http://localhost:8000 and click Re-index Documents to build the FAISS index.

🔧 Changes in This Release

  • Initial public release
  • Fixed: removed non-existent api/ directory reference from README
  • Fixed: removed claude_oauth_provider.py stub reference from project structure
  • Fixed: aligned DEFAULT_EMBEDDING_MODEL across config.py and documentation
  • Fixed: copyright year updated to 2026
  • Security: demand/OpenAI_API_key.pdf excluded via .gitignore
Assets 2
Loading