Self-hosted AI setup for teams requiring complete offline functionality and data privacy

When to Use

Perfect for organizations that need:

Complete data privacy with no external API calls
Offline AI capabilities using local models
Self-hosted infrastructure for sensitive projects
Predictable costs without per-request pricing
Air-gapped environments with no internet access

This configuration provides full AI assistance while keeping all data and processing on your local infrastructure.

Configuration Template

{
  "$schema": "https://commitweave.dev/schema.json",
  "ai": {
    "enabled": true,
    "provider": "local",
    "baseURL": "http://localhost:11434/v1",
    "model": "codellama:7b-instruct",
    "apiKey": "not-required",
    "temperature": 0.4,
    "maxTokens": 100,
    "timeout": 15000,
    "retries": 2,
    "streaming": false,
    "aiSummary": true,
    "contextWindow": 4096
  },
  "localAI": {
    "provider": "ollama",
    "endpoint": "http://localhost:11434",
    "models": {
      "primary": "codellama:7b-instruct",
      "fallback": "llama2:7b-chat", 
      "summary": "codellama:7b-instruct"
    },
    "modelPath": "/opt/ollama/models",
    "gpuAcceleration": true,
    "maxMemory": "8GB",
    "threads": 4
  },
  "commit": {
    "type": {
      "required": true,
      "enum": [
        "feat", "fix", "docs", "style", "refactor",
        "perf", "test", "build", "ci", "chore", "revert"
      ]
    },
    "scope": {
      "required": false,
      "enum": [
        "core", "api", "ui", "cli", "config", "tests",
        "models", "training", "inference", "deployment"
      ]
    },
    "emoji": {
      "enabled": true,
      "style": "conventional"
    },
    "format": {
      "maxLength": 72,
      "minLength": 10,
      "case": "lowercase",
      "wrapBody": 72
    }
  },
  "git": {
    "signoff": false,
    "gpgSign": false,
    "hooks": {
      "skipVerify": false
    }
  },
  "ui": {
    "interactive": true,
    "fancyUI": true,
    "asciiArt": false,
    "animations": true,
    "colors": true,
    "emoji": true,
    "editor": "${EDITOR:-vim}",
    "prompts": {
      "confirmCommit": true,
      "showPreview": true,
      "allowEdit": true,
      "aiProgress": true
    }
  },
  "performance": {
    "caching": true,
    "cacheDir": "~/.commitweave/cache",
    "cacheTTL": 3600,
    "preloadModel": false,
    "batchRequests": false,
    "offlineMode": true
  },
  "privacy": {
    "noTelemetry": true,
    "localOnly": true,
    "encryptCache": true,
    "clearOnExit": false
  }
}

Local AI Setup

Install Ollama (Recommended)

# macOS
brew install ollama

# Linux  
curl -fsSL https://ollama.ai/install.sh | sh

# Windows
# Download from https://ollama.ai/download

Pull Required Models

# Primary coding model
ollama pull codellama:7b-instruct

# Fallback general model  
ollama pull llama2:7b-chat

# Lightweight option for lower-end hardware
ollama pull tinyllama:1.1b-chat

Start Ollama Service

# Start Ollama server
ollama serve

# Verify it's running
curl http://localhost:11434/api/tags

Model Recommendations: codellama:7b-instruct provides excellent commit message generation with ~4GB RAM usage. For resource-constrained environments, use tinyllama:1.1b-chat (~600MB RAM).

Hardware Requirements

Minimum System Requirements

RAM: 4GB available (for 7B models)
Storage: 10GB for models and cache
CPU: 4 cores recommended
GPU: Optional but significantly faster with NVIDIA/AMD GPU

Recommended System Requirements

RAM: 8GB+ available
Storage: 20GB+ SSD storage
CPU: 8+ cores with AVX2 support
GPU: NVIDIA RTX series or AMD RX series

Performance Tuning

// For high-end systems
"localAI": {
  "maxMemory": "16GB",
  "threads": 8,
  "gpuAcceleration": true,
  "model": "codellama:13b-instruct"
}

// For low-end systems  
"localAI": {
  "maxMemory": "2GB", 
  "threads": 2,
  "gpuAcceleration": false,
  "model": "tinyllama:1.1b-chat"
}

Offline Workflow

Daily Development

# AI-powered commits with local processing
commitweave ai --local

# Check model status and performance
commitweave doctor --ai --local

# Generate commit with custom context
commitweave ai --context "Optimizing database queries" --model codellama:7b-instruct

Model Management

# Switch between models
commitweave config --set ai.model llama2:7b-chat

# Check model performance
ollama ps

# Update models
ollama pull codellama:7b-instruct

Alternative Local AI Providers

LM Studio Setup

{
  "ai": {
    "provider": "local",
    "baseURL": "http://localhost:1234/v1",
    "model": "codellama-7b-instruct.Q4_K_M.gguf"
  }
}

LocalAI Setup

{
  "ai": {
    "provider": "local", 
    "baseURL": "http://localhost:8080/v1",
    "model": "codellama-7b-instruct"
  }
}

Custom OpenAI-Compatible API

{
  "ai": {
    "provider": "custom",
    "baseURL": "http://your-internal-api:8000/v1",
    "model": "your-fine-tuned-model",
    "apiKey": "your-internal-key"
  }
}

Troubleshooting

Model Loading Issues

# Check available models
ollama list

# Test model directly
ollama run codellama:7b-instruct "Generate a commit message for adding user authentication"

# Check system resources
ollama ps

Performance Optimization

# Monitor CommitWeave performance
commitweave doctor --ai --performance

# Check cache status
commitweave cache --status

# Clear cache if needed
commitweave cache --clear

Memory Management

Reduce model size: Switch to smaller models like tinyllama
Limit threads: Reduce threads in config to free up CPU
Enable GPU: Use gpuAcceleration: true if available
Adjust memory: Set maxMemory based on available RAM

Benefits vs Tradeoffs

Benefits

✅ Complete data privacy and security
✅ No API costs or rate limits
✅ Works in air-gapped environments
✅ Predictable performance
✅ Customizable models

Tradeoffs

❌ Requires significant local resources
❌ Initial setup complexity
❌ Model quality may vary vs cloud AI
❌ Slower than optimized cloud APIs
❌ Requires ongoing model management

Related Templates: Team Standard • Enterprise Secure

Offline Local AI Configuration