Skip to main content

AI Setup

Configure AI providers and models for msh AI features.

Quick Setup​

msh config ai --provider openai --model gpt-4 --api-key env:OPENAI_API_KEY

Configuration Options​

Provider​

Set the AI provider using --provider:

msh config ai --provider <provider>

Supported Providers:

  • openai - OpenAI (GPT-4, GPT-3.5, etc.)
  • anthropic - Anthropic (Claude 3 Opus, Sonnet, etc.)
  • ollama - Ollama (local models, no API key needed)
  • huggingface - HuggingFace (local models)
  • azure - Azure OpenAI

Model​

Set the model identifier using --model:

msh config ai --model <model>

Examples:

  • gpt-4 - OpenAI GPT-4
  • gpt-3.5-turbo - OpenAI GPT-3.5 Turbo
  • claude-3-opus - Anthropic Claude 3 Opus
  • claude-3-sonnet - Anthropic Claude 3 Sonnet
  • llama2 - Ollama Llama 2

API Key​

Set the API key using --api-key. Use env:VAR_NAME to reference environment variables:

msh config ai --api-key env:OPENAI_API_KEY

Security Best Practice: Always use environment variables, never hardcode API keys.

Custom Endpoint​

For local or custom providers, set a custom endpoint:

msh config ai --endpoint http://localhost:11434/api/generate

Provider-Specific Examples​

OpenAI​

msh config ai \
  --provider openai \
  --model gpt-4 \
  --api-key env:OPENAI_API_KEY

Environment Variable:

# .env
OPENAI_API_KEY=sk-...

Anthropic​

msh config ai \
  --provider anthropic \
  --model claude-3-opus \
  --api-key env:ANTHROPIC_API_KEY

Environment Variable:

# .env
ANTHROPIC_API_KEY=sk-ant-...

Ollama (Local)​

msh config ai \
  --provider ollama \
  --model llama2 \
  --endpoint http://localhost:11434/api/generate

Note: No API key needed for Ollama.

Azure OpenAI​

msh config ai \
  --provider azure \
  --model gpt-4 \
  --api-key env:AZURE_OPENAI_API_KEY \
  --endpoint https://your-resource.openai.azure.com

Configuration Storage​

AI configuration is stored in:

  • Global: ~/.msh/config (applies to all projects)
  • Project-level: msh.yaml (project-specific override)

Project-level configuration takes precedence over global configuration.

Verifying Configuration​

After configuration, verify it works:

# Generate manifest (required before AI commands)
msh manifest

# Test AI command
msh ai explain assets/revenue.msh

Troubleshooting​

"AI configuration incomplete"​

Issue: AI commands fail with configuration error.

Solution:

  1. Check configuration: cat ~/.msh/config or check msh.yaml
  2. Verify API key is set: echo $OPENAI_API_KEY
  3. Reconfigure if needed: msh config ai --provider ...

"API key not found"​

Issue: Environment variable not set.

Solution:

  1. Set environment variable: export OPENAI_API_KEY=sk-...
  2. Or add to .env file: OPENAI_API_KEY=sk-...
  3. Verify: echo $OPENAI_API_KEY

"Provider not supported"​

Issue: Provider name incorrect.

Solution:

  • Check provider name spelling
  • Use one of: openai, anthropic, ollama, huggingface, azure

Next Steps​