Grafana Agent PoC with Semantic Kernel

This project demonstrates a proof of concept for building an AI agent using the Semantic Kernel Python SDK that interfaces with Grafana through a local MCP server.

Prerequisites

• Python 3.8+
• Docker and Docker Compose
• Ollama

Setup

1. Create a virtual environment and activate it:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

2. Install dependencies:

pip install -r requirements.txt

3. Configure environment variables:

   • Copy the .env.example file to create a new .env file:

   cp .env.example .env
   

   • Update the .env file with your credentials:
     • AZURE_OPENAI_DEPLOYMENT_NAME: Your Azure OpenAI deployment name
     • AZURE_OPENAI_API_VERSION: API version (e.g., 2024-12-01-preview)
     • AZURE_OPENAI_API_KEY: Your Azure OpenAI API key
     • AZURE_OPENAI_BASE_URL: Your Azure OpenAI base URL
     • GRAFANA_MCP_URL: Local MCP server URL (default: http://localhost:8000/sse)
     • GRAFANA_URL: Grafana instance URL (default: http://localhost:3000)
     • GRAFANA_ADMIN_USERNAME: Grafana admin username
     • GRAFANA_ADMIN_PASSWORD: Grafana admin password
     • GRAFANA_API_KEY: Your Grafana API key (create one in Grafana UI)

4. Start Grafana using Docker Compose:

cd docker
docker-compose up -d

5. Configure Grafana:
   • Access Grafana at http://localhost:3000
   • Log in with the credentials set in your .env file
   • Go to Configuration -> API Keys
   • Create a new API key with Admin role
   • Copy the generated key to your .env file's GRAFANA_API_KEY

Running the Applications

The project provides three ways to interact with Grafana:

1. Basic Console Interface (src/console.py):

   python src/console.py
   

   Basic chat based console interface powered by llm and Semantic Kernel.

2. Interactive Console Agent (src/console_agent.py):

   python src/console_agent.py
   

   Agent based console interface powered by llm and Semantic Kernel, allowing for more complex interactions.

3. REST API Server (src/main.py):

   PYTHONPATH=. uvicorn src.main:app --reload --host 0.0.0.0 --port 8080 --log-level debug
   

   Access the API at http://localhost:8080

   • REST API documentation: http://localhost:8080/docs
   • Health check: http://localhost:8080/health

API Endpoints

Health Check

GET /health

Query the Agent

POST /query
Content-Type: application/json

{
    "query": "Your natural language query here"
}

Grafana Access

Grafana UI is available at http://localhost:3000

• Use the credentials specified in your .env file

Example Usage

Here are some example queries you can send to the agent:

1. Create a new dashboard:

{
    "query": "Create a new dashboard called 'System Metrics'"
}

2. Get dashboard information:

{
    "query": "Show me the details of dashboard with UID xyz123"
}

3. Update a dashboard:

{
    "query": "Update the System Metrics dashboard to include a CPU usage panel"
}

Project Structure

graphana-agent-poc/
├── docker/
│   └── docker-compose.yml       # Grafana container configuration
├── src/
│   ├── agent/
│   │   ├── plugins/
│   │   │   └── grafana_plugin.py    # Semantic Kernel plugin for Grafana
│   │   └── skills/
│   │       └── grafana_skills.py     # Implementation of Grafana operations
│   └── main.py                       # FastAPI application and agent setup
├── requirements.txt                   # Python dependencies
└── README.md                         # This file