Proxmox LangChain Agent

A lightweight FastAPI service that provides a chat endpoint backed by LangChain and Redis-based message history. This service uses an Ollama LLM for responses and persists conversations in Redis. It is designed to interact with a Proxmox MCP (Model Context Protocol) server, allowing conversational access to Proxmox cluster and VM information.

Repository Structure

agent-proxmox/
├── Dockerfile
├── app.py
├── requirements.txt
└── README.md

• Dockerfile: Builds a slim Python image with the application and dependencies.
• app.py: FastAPI application defining a /chat endpoint that runs a LangChain agent with Proxmox-specific tools and Redis-backed history.
• requirements.txt: Python dependencies needed to run the agent.

Prerequisites

• Docker Engine (20.x+)
• Redis instance accessible by the agent
• Proxmox MCP server accessible by the agent (see below)
• (Optional) Docker Compose if integrating into a larger stack

Proxmox MCP Server Integration

This agent is designed to work with a Proxmox MCP (Model Context Protocol) server. The MCP server exposes HTTP endpoints for cluster and VM information, which the agent accesses using custom LangChain tools:

• get_vm_list: Calls the MCP endpoint /mcp/context/vms to retrieve a list of all VMs and their statuses.
• get_cluster_info: Calls the MCP endpoint /mcp/context/cluster to retrieve high-level Proxmox cluster status info.

You must have a running MCP server accessible at the address configured in app.py (default: http://mcp:8008).

Environment Variables

Variable	Default	Description
REDIS_HOST	localhost	Hostname or IP of your Redis server
REDIS_PORT	6379	Port number of your Redis server
REDIS_PASSWORD	none	Password for Redis (required if Redis is secured)
LANGCHAIN_LOG_DIR	./logs	Directory path (inside container) for logs

Make sure to set REDIS_PASSWORD if your Redis instance requires authentication.

Building the Docker Image

From within the project directory:

docker build -t agent-proxmox .

Running the Container

A minimal docker run example:

docker run -d \
  --name agent-proxmox \
  -p 8501:8501 \
  -e REDIS_HOST=ai-redis \
  -e REDIS_PORT=6379 \
  -e REDIS_PASSWORD=$REDIS_PASSWORD \
  -e LANGCHAIN_LOG_DIR=/logs \
  -v $(pwd)/logs:/logs \
  agent-proxmox

This exposes the FastAPI app on port 8501 and mounts a local logs/ directory for persistent logging.

Integrating with Docker Compose

If you have a larger Docker Compose setup, add this service:

services:
  agent-proxmox:
    build: ./agent-proxmox
    container_name: agent-proxmox
    networks:
      - ai-stack
    depends_on:
      - ai-redis
      - mcp
    environment:
      - REDIS_HOST=ai-redis
      - REDIS_PORT=6379
      - REDIS_PASSWORD=${REDIS_PASSWORD}
      - LANGCHAIN_LOG_DIR=${LANGCHAIN_LOG_DIR}
    volumes:
      - ./agent-proxmox/logs:${LANGCHAIN_LOG_DIR}
    ports:
      - "8501:8501"

API Usage

Send a POST request to /chat with a JSON body { "message": "Your question here" }:

curl -X POST http://localhost:8501/chat \
  -H "Content-Type: application/json" \
  -d '{"message":"Show me all VMs in the cluster"}'

Response:

{ "response": "<LLM reply>" }

Logs

Conversation logs are written to langchain_YYYYMMDD.log in the LANGCHAIN_LOG_DIR directory. Adjust verbosity by modifying the logging.basicConfig settings in app.py.

CI/CD Pipeline (GitLab)

This project uses a GitLab CI/CD pipeline to automate validation, linting, building, and deployment of the Docker container to the GitLab Container Registry. The pipeline is defined in .gitlab-ci.yml and includes the following stages:

• validate: Installs dependencies and checks Python syntax.
• lint: Runs flake8 to enforce code style and quality.
• deploy: Builds and pushes the Docker image to the registry (only on main branch and tags).

The pipeline uses GitLab's built-in CI/CD variables for authentication and registry information. No additional variables are required for standard operation.

FastAPI Application Overview

• The main application is in app.py and uses FastAPI to provide a /chat endpoint.
• The endpoint expects a POST request with a JSON body: { "message": "Your question here" }.
• The backend uses LangChain's agent framework with two custom tools for Proxmox MCP integration.
• Conversation history is stored in Redis using langchain_community.chat_message_histories.RedisChatMessageHistory.
• The LLM is provided by an Ollama server (default: http://ollama:11434).
• Logging is configured to write conversation logs to the directory specified by LANGCHAIN_LOG_DIR (default: ./logs).

Developer Notes

• Code Style: The project enforces PEP8 compliance using flake8. Ensure your code passes flake8 app.py before committing.
• Environment Variables: See the table above. You must set REDIS_PASSWORD if your Redis instance is secured.
• Testing the API: Use the provided curl example or a tool like Postman to interact with the /chat endpoint.
• Extending the Agent: To add new tools, define a new function and add it to the tools list in app.py using the Tool class from LangChain.
• Pipeline Troubleshooting: If the Docker image fails to push, ensure you are pushing from the main branch or a tag, as the pipeline only deploys in those cases.
• Logs: All conversations are logged with timestamps. Check the logs/ directory for daily log files.

Contributing

1. Fork the repository and create a feature branch.
2. Ensure your code passes linting and validation locally:

   pip install -r requirements.txt
   flake8 app.py
   python -m py_compile app.py
   

3. Push your branch and create a merge request.

---

For questions or issues, please open an issue in the GitLab repository.

Happy chaining!