Gemini_MCP

# Claude MCP

A multi-agent AI assistant powered by Gemini LLM using the ReAct pattern to orchestrate specialized agents through Model Context Protocol (MCP). Features a LangGraph supervisor with streaming Streamlit and Gradio UIs.

## ✨ Features

- **ReAct-based supervisor** – dynamically routes queries to specialized agents
- **5 Specialized Agents** – crypto, stocks, portfolio tracking, RAG document Q&A, web search
- **Real-time streaming** – see AI reasoning steps as they happen
- **Document upload** – chat with PDFs, TXT, and DOCX files
- **Cloud integration** – ChromaDB Cloud for vectors, Google Cloud SQL for portfolio data
- **Docker support** – containerized MCP servers

## 🛠️ Tech Stack

- **AI/LLM**: Google Gemini 2.5 Pro, LangChain, LangGraph
- **UI**: Streamlit (main), Gradio 5.0+ (for demos)
- **Agent Protocol**: FastMCP (Model Context Protocol)
- **Databases**: ChromaDB Cloud (vectors), Google Cloud SQL (PostgreSQL)
- **APIs**: Alpha Vantage, CoinGecko, DuckDuckGo
- **DevOps**: Docker, Docker Compose


## ⚙️ How It Works

The ReAct supervisor uses a reasoning loop (max 5 steps):
1. **Think** – Analyze query and determine needed information
2. **Act** – Call appropriate agent(s)
3. **Observe** – Process agent responses
4. **Repeat** – Continue until sufficient information gathered
5. **Synthesize** – Generate final answer


## 🚀 Quick Start
```bash
# Clone and install
git clone <repository-url>
cd Gemini_MCP
pip install -r requirements.txt
```

Create `.env` file:
```bash
GOOGLE_API_KEY=your_google_ai_api_key
CHROMA_API_KEY=your_chroma_api_key
CHROMA_TENANT=your_tenant_id
CHROMA_DATABASE=your_database_name
ALPHA_VANTAGE_API_KEY=your_alpha_vantage_key
COINGECKO_API_KEY=your_coingecko_key
GCP_PROJECT_ID=your_project_id
CLOUD_SQL_INSTANCE_CONNECTION_NAME=project:region:instance
CLOUD_SQL_DB_NAME=finance_tracker
CLOUD_SQL_DB_USER=your_db_user
CLOUD_SQL_DB_PASS=your_db_password
```

Run database schema:
```bash
psql -h <cloud-sql-host> -U <user> -d finance