🇮🇳 India Insight Agent

An AI-powered analytical engine that connects directly to the Ministry of Statistics and Programme Implementation (MoSPI) via the Model Model Context Protocol (MCP). This agent fetches live development data and provides deep, data-driven insights into India's growth sectors.

---

🚀 Key Features

• Live Data Retrieval: Connects to the official mcp.mospi.gov.in endpoint via SSE.
• Sequential Reasoning: Uses a two-stage "Researcher → Analyst" flow to ensure accuracy before interpretation.
• Cloud Native: Fully containerized with a multi-stage Docker build, optimized for Google Cloud Run.
• Built with ADK: Leverages Google's Agent Development Kit for seamless LLM orchestration.

---

📁 Project Structure

india-insight-agent/
├── src/
│   ├── main.py             # App entry point (ADK Web/API)
│   ├── agent/
│   │   ├── insight_agent.py # Core logic: SequentialAgent definition
│   │   └── prompts.py      # Specialized System Prompts for MoSPI
│   └── tools/
│       └── mospi_client.py  # MCP SSE Client connection logic
├── Dockerfile              # Optimized multi-stage build using 'uv'
├── pyproject.toml          # Dependency and project configuration
└── .env                    # Local environment variables

---

🛠️ Tech Stack

Component	Technology
Language	Python 3.12
Package Manager	uv
Agent Framework	Google ADK
Protocol	MCP (Model Context Protocol)
Infrastructure	Google Cloud Run

---

⚙️ Local Setup

1. Install uv

If you don't have it yet, install the faster Python package manager:

curl -LsSf https://astral.sh/uv/install.sh | sh

2. Initialize and Install

# Sync dependencies from pyproject.toml
uv sync

3. Configure Environment

Create a .env file in the root directory:

GOOGLE_API_KEY=your_gemini_api_key
PORT=8080

4. Run the Agent

uv run src/main.py

Access the interface at http://localhost:8080.

---

☁️ Deployment to Cloud Run

Deploy directly from your source code. Cloud Run will use the Dockerfile to build the image automatically.

gcloud run deploy india-insight-agent \
  --source . \
  --region asia-south1 \
  --allow-unauthenticated \
  --set-env-vars GOOGLE_API_KEY=your_gemini_key_here \
  --max-instances 3

---

🤖 Agent Logic

The IndiaInsightAgent follows a strict Fetch-then-Analyze pattern:

1. DataResearcher: Uses MoSPI tools to find indicators (GDP, CPI, Energy, etc.) and fetches the latest 5-10 years of raw data.
2. TrendAnalyst: Receives the raw JSON, calculates growth rates (CAGR), and interprets the data within the context of India's development goals.

---

📄 License

Distributed under the MIT License. See LICENSE for more information.