AI Providers

Connect ABOV3 to multiple AI providers for maximum flexibility and model selection.

🚀 Quick Setup

Most providers work out of the box. Just run ./abov3-linux-x64 auth login for Anthropic OAuth, or add API keys to your configuration.

Supported Providers

🤖 Anthropic (Claude)

Recommended

Models: Claude 3 Opus, Sonnet, Haiku

Authentication: OAuth (Claude Pro/Max) or API Key

OAuth Setup (Recommended)

# Login with your Claude Pro/Max account
./abov3-linux-x64 auth login

# This will:
# 1. Open browser for authentication
# 2. Save credentials securely
# 3. Auto-refresh tokens as needed

API Key Setup

# Add to ~/.config/abov3/abov3.json
{
  "provider": {
    "anthropic": {
      "apiKey": "sk-ant-api03-..."
    }
  }
}

🏠 Ollama (Local Models)

Free & Private

Models: Llama, Mistral, Qwen, CodeLlama, and more

Requirements: Ollama installed and running locally

Setup

# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# Start Ollama service
ollama serve

# Pull models you want
ollama pull qwen2.5-coder:14b
ollama pull mistral:7b
ollama pull llama3:8b

# Use in ABOV3
./abov3-linux-x64 run -m ollama/qwen2.5-coder:14b "Your prompt"

Remote Ollama

# Connect to remote Ollama instance
export OLLAMA_BASE_URL="http://192.168.1.100:11434"

# Or add to config
{
  "provider": {
    "ollama": {
      "baseUrl": "http://192.168.1.100:11434"
    }
  }
}

🌐 OpenAI (GPT)

Models: GPT-4, GPT-4 Turbo, GPT-3.5 Turbo

Authentication: API Key required

Setup

# Add to ~/.config/abov3/abov3.json
{
  "provider": {
    "openai": {
      "apiKey": "sk-..."
    }
  }
}

# Or use environment variable
export OPENAI_API_KEY="sk-..."

🔷 Azure OpenAI

Models: GPT-4, GPT-3.5 (via Azure deployment)

Authentication: Azure credentials

Setup

{
  "provider": {
    "azure": {
      "apiKey": "your-azure-key",
      "endpoint": "https://your-resource.openai.azure.com",
      "deployment": "gpt-4"
    }
  }
}

🌊 Google (Gemini)

Beta

Models: Gemini Pro, Gemini Ultra

Authentication: API Key

Setup

{
  "provider": {
    "google": {
      "apiKey": "your-gemini-api-key"
    }
  }
}

Provider Configuration

Global Configuration

Edit ~/.config/abov3/abov3.json:

{
  "defaultModel": "anthropic/claude-3-sonnet",
  "provider": {
    "anthropic": {
      "apiKey": "sk-ant-...",
      "maxTokens": 4096,
      "temperature": 0.7
    },
    "openai": {
      "apiKey": "sk-...",
      "organization": "org-..."
    },
    "ollama": {
      "baseUrl": "http://localhost:11434"
    }
  }
}

Environment Variables

# Anthropic
export ANTHROPIC_API_KEY="sk-ant-..."

# OpenAI
export OPENAI_API_KEY="sk-..."

# Ollama (if not localhost)
export OLLAMA_BASE_URL="http://192.168.1.100:11434"

# Azure
export AZURE_OPENAI_KEY="..."
export AZURE_OPENAI_ENDPOINT="https://..."

# Google
export GOOGLE_API_KEY="..."

Model Comparison

Provider	Model	Context	Speed	Cost	Best For
Anthropic	Claude 3 Opus	200K	Medium	$$$	Complex reasoning, large codebases
Anthropic	Claude 3 Sonnet	200K	Fast	$$	Balanced performance
Anthropic	Claude 3 Haiku	200K	Very Fast	$	Quick tasks, high volume
OpenAI	GPT-4	128K	Medium	$$$	General purpose, creative tasks
OpenAI	GPT-3.5 Turbo	16K	Very Fast	$	Simple tasks, prototyping
Ollama	Qwen2.5-Coder:14b	32K	Fast*	Free	Code generation, local development
Ollama	Mistral:7b	8K	Very Fast*	Free	General purpose, privacy-sensitive

* Speed depends on local hardware

Managing Providers

List Available Models

# Show all models from configured providers
./abov3-linux-x64 models

# Output includes:
# - Provider name
# - Model ID
# - Availability status

Switch Between Providers

# Use specific model for a session
./abov3-linux-x64 -m anthropic/claude-3-opus

# One-off with different model
./abov3-linux-x64 run -m ollama/mistral:7b "Your prompt"

# Set default in config
{
  "defaultModel": "anthropic/claude-3-sonnet"
}

Test Provider Connection

# Test with simple prompt
./abov3-linux-x64 run --print-logs "Hello, are you working?"

# Check authentication status
./abov3-linux-x64 auth list

💡 Provider Tips

Anthropic OAuth: Best for Claude Pro/Max users - no API costs
Ollama: Perfect for offline work and privacy-sensitive projects
Multiple Providers: Configure all providers for maximum flexibility
Model Selection: Use faster models for simple tasks, powerful models for complex work
Cost Management: Monitor API usage and use local models when possible

Troubleshooting

Authentication Failed

# Check credentials
./abov3-linux-x64 auth list

# Re-authenticate Anthropic
./abov3-linux-x64 auth logout
./abov3-linux-x64 auth login

# Verify API keys in config
cat ~/.config/abov3/abov3.json | grep apiKey

Ollama Connection Issues

# Check if Ollama is running
curl http://localhost:11434/api/tags

# Restart Ollama
systemctl restart ollama  # Linux with systemd
ollama serve              # Manual start

# Check available models
ollama list

Rate Limiting

If you encounter rate limits:

Switch to a different provider temporarily
Use local Ollama models for development
Implement retry logic in scripts
Consider upgrading your API plan