RAG Financial Analysis System 🤖📊

A real Retrieval-Augmented Generation (RAG) pipeline for financial forecasting — powered by Pinecone vector search, live NewsData.io market intelligence, Gemini AI, and pandas data analysis.

Upload your own CSV/Excel data → get AI-powered revenue forecasts with interactive charts and PDF reports.

---

✨ Features

Feature	Description
📂 CSV/Excel Upload	Upload your own financial data via Telegram or the desktop GUI
🔍 Real Pinecone Search	Semantic vector search with Gemini text-embedding-004
📰 Live News	Real-time financial news from NewsData.io
🧠 AI Forecasting	Gemini 2.0 Flash-Lite with chain-of-thought reasoning
📊 Charts	Dark-themed Matplotlib revenue charts with confidence intervals
📄 PDF Reports	Professional multi-page executive report export
🖥️ Desktop GUI	Tkinter dashboard with live pipeline logs and data preview

---

⚡ Quick Start

1. Install Dependencies

pip install -r requirements.txt

2. Configure API Keys

Copy .env.example to .env and fill in your keys:

cp .env.example .env

Key	Source	Required
TELEGRAM_BOT_TOKEN	@BotFather	✅ Yes
GEMINI_API_KEY	aistudio.google.com	✅ Yes
PINECONE_API_KEY	app.pinecone.io	Optional
NEWSDATA_API_KEY	newsdata.io	Optional
OPENROUTER_API_KEY	openrouter.ai	Optional (fallback)

3. Run

Telegram Bot only:

python rag_demo.py

Desktop GUI only:

python gui_app.py

Both together:

python launch_all.py

---

🤖 Bot Commands

Command	Action
/start	Welcome message
/demo	Run sample forecast with built-in demo data
/report	Generate PDF executive report of last forecast
/help	Commands and example queries
Send CSV/Excel	Upload and auto-analyze your data
Any text	Run RAG pipeline → AI forecast

---

📊 How It Works

User uploads CSV/Excel or sends query
    ↓
[Node 1/7] pandas data analysis (real statistics)
    ↓
[Node 2/7] NewsData.io live financial news
    ↓
[Node 3/7] Pinecone vector search (768-dim embeddings)
    ↓
[Node 4/7] Gemini text-embedding-004 generation
    ↓
[Node 5/7] Context assembly (data + news + docs)
    ↓
[Node 6/7] Gemini 2.0 Flash-Lite AI forecast
    ↓
[Node 7/7] Format response + chart + PDF

---

📁 Project Structure

rag-financial-analysis/
├── rag_demo.py              ← Telegram bot (slim handler-only entry point)
├── gui_app.py               ← Desktop GUI dashboard
├── launch_all.py            ← Dual launcher (bot + GUI)
├── core/
│   ├── __init__.py
│   ├── analyzer.py          ← Real pandas data analysis engine
│   ├── ai_engine.py         ← Gemini AI + embeddings + fallback chain
│   ├── chart_engine.py      ← Dark-themed Matplotlib charts
│   ├── pipeline.py          ← RAG pipeline orchestrator
│   ├── formatters.py        ← Telegram HTML message formatters
│   └── pdf_report.py        ← Executive PDF report generator
├── data/
│   ├── __init__.py
│   ├── sample_data.py       ← Demo datasets + fallback constants
│   └── samples/
│       └── demo_sales.csv   ← 12-month sample dataset
├── requirements.txt
├── .env.example
├── .gitignore
├── Procfile
└── run.bat

---

⚙️ Tech Stack

Component	Technology
Bot Framework	python-telegram-bot 21+
AI Model	Google Gemini 2.0 Flash-Lite
Embeddings	Gemini text-embedding-004 (768-dim)
Vector DB	Pinecone (serverless)
Live News	NewsData.io API
Data Analysis	pandas + numpy
Charts	Matplotlib
PDF Export	ReportLab
Desktop GUI	Tkinter
Fallback AI	OpenRouter

---

📄 License

MIT