MCP Server with OpenAI, Git, Filesystem, and Prometheus Integration

This repository contains a Model Control Plane (MCP) server implementation that supports OpenAI services, Git repository analysis, local filesystem operations, and Prometheus integration.

Project Structure

Requirements

• Python 3.8+

• FastAPI

• Uvicorn

• OpenAI SDK

• GitPython

• Requests

• Docker and Docker Compose (for Prometheus features)

Installation

1. Clone this repository

2. Install the dependencies:

Environment Variables

Set the following environment variables:

For Azure OpenAI:

For Standard OpenAI:

For Prometheus:

Running the Server

Start the MCP server:

Or for more options:

The server will be available at http://localhost:8000.

Unified Testing Tool

We provide a unified testing script that gives you a user-friendly interface to all testing functionality:

This interactive script provides:

• Filesystem tests

• Git integration tests

• Memory analysis tools

• Prometheus tests & memory stress

• MCP server management

• Environment setup

Individual Tests

You can also run individual tests directly:

Test the OpenAI integration:

Test the Git integration (provide a Git repository URL):

Test the Git diff functionality (analyze requirements compatibility):

Test the filesystem functionality:

Test the langflow integration with MCP:

Test the Prometheus integration:

Advanced Git Analysis

For more advanced Git repository analysis with AI recommendations:

You can also search for specific patterns in the repository:

Or analyze the last commit diff with AI insights:

Memory Analysis Tools

MCP includes several tools for memory monitoring and analysis:

You can also simulate memory pressure for testing:

Prometheus Integration

Setup

1. Start the Prometheus stack using Docker Compose:

This will start:

• Prometheus server (accessible at http://localhost:9090)

• Node Exporter (for host metrics)

• cAdvisor (for container metrics)

2. For stress testing, you can start the memory stress container:

Or use the container test script:

Docker Configuration and Reset Scripts

This project includes multiple Docker configurations and reset scripts for reliable operation across different environments:

Docker Configurations

• Standard Configuration (docker-compose.yml): Uses custom Dockerfiles for Prometheus and Langflow to ensure consistent permissions across systems.

• Bridge Network Configuration (docker-compose.bridge.yml): Alternative configuration that uses bridge networking for environments where host networking is problematic.

Custom Dockerfiles for Solving Permission Issues

The project uses custom Dockerfiles for both Prometheus and Langflow to solve common permission issues:

• Dockerfile.prometheus: Sets up the Prometheus configuration with proper permissions for the nobody user.

• Dockerfile.langflow: Copies the components directory into the container without changing file ownership, allowing Langflow to access the components without permission errors.

This approach eliminates the need for volume mounts that can lead to permission conflicts across different machines and user configurations.

Reset Scripts

• All Services Reset (reset-all.sh): Reset all containers with a single command.

  # Basic reset (rebuilds containers with existing volumes) ./reset-all.sh # Full reset (removes volumes and rebuilds containers) ./reset-all.sh --clean

• Individual Service Reset:

  # Reset only Prometheus ./reset-prometheus.sh # Reset only Langflow ./reset-langflow.sh

These scripts ensure that the containers are properly configured with correct permissions and the latest code changes.

Troubleshooting

If you encounter permission issues:

1. Use the reset scripts to rebuild the containers

2. Check the logs with docker compose logs <service_name>

3. Make sure any components added to Langflow are included in the Dockerfile.langflow

Cross-Machine Deployment

When deploying to a new machine:

1. Clone the repository

2. Make reset scripts executable: chmod +x *.sh

3. Run the reset script: ./reset-all.sh

The custom Dockerfiles automatically handle all permission issues that might occur across different systems.

Using Prometheus Client

The MCPAIComponent class includes Prometheus capabilities:

Useful PromQL Queries

• CPU Usage: rate(node_cpu_seconds_total{mode!="idle"}[1m])

• Memory Usage: node_memory_MemTotal_bytes - node_memory_MemAvailable_bytes

• Disk Usage: node_filesystem_avail_bytes{mountpoint="/"} / node_filesystem_size_bytes{mountpoint="/"}

• Container CPU Usage: rate(container_cpu_usage_seconds_total[1m])

• Container Memory Usage: container_memory_usage_bytes

API Endpoints

OpenAI Endpoints

• GET /v1/models - List all available models

• GET /v1/models/{model_id} - Get information about a specific model

• POST /v1/models/azure-gpt-4/completion - Generate text completion using Azure OpenAI

• POST /v1/models/azure-gpt-4/chat - Generate chat response using Azure OpenAI

• POST /v1/models/openai-gpt-chat/chat - Generate chat response using OpenAI chat model

• POST /v1/models/openai-gpt-completion/completion - Generate text completion using OpenAI completion model

Git Integration Endpoints

• POST /v1/models/git-analyzer/analyze - Analyze a Git repository

• POST /v1/models/git-analyzer/search - Search a Git repository for files matching a pattern

• POST /v1/models/git-analyzer/diff - Get the diff of the last commit in a repository

Filesystem Endpoints

• POST /v1/models/filesystem/list - List contents of a directory

• POST /v1/models/filesystem/read - Read a file's contents

• POST /v1/models/filesystem/read-multiple - Read multiple files at once

• POST /v1/models/filesystem/write - Write content to a file

• POST /v1/models/filesystem/edit - Edit a file with multiple replacements

• POST /v1/models/filesystem/mkdir - Create a directory

• POST /v1/models/filesystem/move - Move a file or directory

• POST /v1/models/filesystem/search - Search for files matching a pattern

• POST /v1/models/filesystem/info - Get information about a file or directory

Prometheus Endpoints

• POST /v1/models/prometheus/query - Execute an instant query

• POST /v1/models/prometheus/query_range - Execute a range query

• POST /v1/models/prometheus/series - Get series data

• GET /v1/models/prometheus/labels - Get all available labels

• POST /v1/models/prometheus/label_values - Get values for a specific label

• GET /v1/models/prometheus/targets - Get all targets

• GET /v1/models/prometheus/rules - Get all rules

• GET /v1/models/prometheus/alerts - Get all alerts

Client Usage

You can use the MCPAIComponent in your LangFlow pipelines by providing the MCP server URL:

Using the GitCodeAnalyzer Class

For more structured Git analysis, you can use the GitCodeAnalyzer class:

Troubleshooting

Prometheus Issues

1. Verify Prometheus is running: docker ps | grep prometheus

2. Check you can access the Prometheus UI: http://localhost:9090

3. Verify the MCP server is running and accessible

4. Check the MCP server logs for errors

5. Try simple queries first to verify connectivity (e.g., up query)

OpenAI Issues

1. Verify your API keys are set correctly

2. Check for rate limiting or quota issues

3. Verify you're using supported models for your API key

Git Issues

1. Ensure the Git repository URL is accessible

2. Check for authentication issues if using private repositories

3. Ensure GitPython is installed correctly