FinSightAI couples a modern React + Node.js dashboard with a Python‑powered AI/ML pipeline so you can ingest PDFs, index their contents, and ask natural‑language questions against your documents. The frontend provides real‑time charts and an intuitive interface, while the backend runs a Retrieval‑Augmented Generation (RAG) workflow built with LangChain, FAISS, and Ollama for lightning‑fast document search and answer generation.
| Tool | Purpose |
|---|---|
| React | Component‑driven UI |
| JavaScript | Application logic & state management |
| CSS / Tailwind | Responsive styling |
| VS Code | Primary IDE with ESLint & Prettier |
- Exposes REST endpoints for the React client
- Proxies requests to the Python AI/ML service
| Library / Module | Role |
|---|---|
os, warnings, re |
File handling, logging, regex |
| pandas | Tabular data manipulation |
pdfminer.six (extract_text) |
PDF → text extraction |
| LangChain core | Orchestrates RAG |
• Document, RecursiveCharacterTextSplitter |
Wrap text & chunk large files |
| • LangChain Community FAISS | In‑memory vector store |
| • HuggingFaceEmbeddings | Generate vector embeddings |
• OllamaLLM |
Local LLM inference |
• RetrievalQA |
Retrieval + generation chain |
- FAISS: Similarity‑search index
- MongoDB & Mongoose (optional): Metadata / user sessions
| Tool | Purpose |
|---|---|
| Postman | API testing & exploration |
| nodemon | Auto‑reload Node server in dev |
Prerequisites
– Python 3.8 + (withpip)
– Node.js 16 + andnpm
– MongoDB (optional, for metadata storage)
git clone https://github.com/YourUser/FinSightAI.git
cd FinSightAIcd python-service
pip install -r requirements.txtpython ingest_pdfs.py --pdf-dir ./docsLaunch the FastAPI / Flask server (defaults to http://localhost:8000)
python app.pycd backend
npm installcp .env.example .env#4. Start the React frontend (terminal #3)
cd finsightai-client
npm install
npm start # opens http://localhost:3000FinSightAI transforms static PDFs into living data— searchable, analyzable, and actionable—by uniting a React + Node.js dashboard with a Python‑based RAG engine. With rapid vector search (FAISS), local LLM inference (Ollama), and an ergonomic UI, the project showcases how full‑stack AI can turn raw documents into instant insight.
This is only the beginning. Future milestones include:
- Real‑time WebSocket streaming for live market and sentiment data
- Drag‑and‑drop widgets so users can build custom dashboards
- Role‑based authentication and shared workspaces for team collaboration
- Mobile clients via React Native
- Alerting & scheduled reports delivered via email or push notifications
We welcome issues, feature requests, and pull requests—check out CONTRIBUTING.md to get started. Together we can evolve FinSightAI into a best‑in‑class, open‑source platform for document‑driven intelligence.