Document Analysis App with Docker Model Runner

This application allows you to upload PDF documents, extract text, and analyze the content using a locally running LLM via Docker Model Runner.

Prerequisites

• Docker Desktop (version 4.40 or later) with Docker Model Runner enabled
• The Llama 3 model pulled locally (docker model pull ai/llama3.2:1B-Q8_0)

Setup Instructions

1. Enable Docker Model Runner in Docker Desktop

   docker desktop enable model-runner
   

2. Pull the Llama 3 model

   docker model pull ai/llama3.2:1B-Q8_0
   

3. Start the application

   docker-compose up -d
   

Accessing the Application

• Document Analysis App: http://localhost:8080
  • Upload PDF documents for AI-powered analysis
  • View extracted insights and summaries

Monitoring and Observability

The application includes a comprehensive observability stack:

• Jaeger UI (Distributed Tracing): http://localhost:16686

  • Monitor request flows across services
  • Analyze performance bottlenecks

• Prometheus (Metrics): http://localhost:9090

  • Query system and application metrics
  • Monitor resource usage and performance

• Grafana (Dashboards): http://localhost:3001

  • View preconfigured dashboards
  • Default credentials: admin/admin

Application Architecture

This application consists of:

• Frontend: React application with document upload capability (port 8080)
• Backend (MCP Server): Express.js server that extracts text from PDFs and calls Docker Model Runner (port 3000)
• Docker Model Runner: Docker Desktop extension that runs the Llama 3 model locally
• Observability Stack: OpenTelemetry, Jaeger, Prometheus, and Grafana for monitoring

User Guide

1. Access the application at http://localhost:8080
2. Drag and drop a PDF file onto the upload area, or click "Browse Files"
3. Wait for the analysis to complete
4. Review the generated insights, summary, and key points

Note on Docker Model Runner

If Docker Model Runner is not correctly set up or the specified model is not available, the application will still function in fallback mode:

• PDF text extraction will work
• A basic analysis will be provided without LLM enhancement
• Error details will be displayed in the analysis result

To ensure full AI functionality, verify that:

• Docker Model Runner is enabled in Docker Desktop
• The correct model is pulled and available
• The Docker host can reach the Model Runner API (typically at http://host.docker.internal:12434/v1)

Troubleshooting

• If you're unable to access the application, check if all containers are running with docker-compose ps
• View logs for specific services with docker-compose logs [service-name] (e.g., docker-compose logs frontend)
• Ensure Docker Model Runner is enabled in Docker Desktop settings
• If document analysis fails, check backend logs with docker-compose logs mcp-server

License

MIT