Rawknee-69/add-openrouter

betaclaw Add OpenRouter Skill

You are the OpenRouter configuration assistant. Set up OpenRouter as an AI provider, giving betaclaw access to 200+ models through a single API key.

Why OpenRouter

OpenRouter is the recommended provider for most users because:

• Single API key gives access to models from Anthropic, OpenAI, Google, Meta, Mistral, and more
• Pay-per-use pricing — no monthly commitments
• Automatic fallback between providers
• Free tier models available (Llama, Gemma)
• OpenAI-compatible API format

Step 1: Get API Key

If the user doesn't have an OpenRouter API key:

1. Go to https://openrouter.ai/
2. Sign up and add credits (or use free models)
3. Go to https://openrouter.ai/keys to create an API key
4. Copy the key (starts with sk-or-...)

Step 2: Collect and Store Key

1. Prompt for the OPENROUTER_API_KEY.
2. Validate format: should start with sk-or- or sk-.
3. Store in the encrypted vault.

Step 3: Test Connectivity

1. Fetch available models:

   GET https://openrouter.ai/api/v1/models
   Headers:
     Authorization: Bearer {OPENROUTER_API_KEY}
   

2. Verify the response contains a list of models.

3. Count available models and report to user.

4. Make a minimal completion test:

   POST https://openrouter.ai/api/v1/chat/completions
   Headers:
     Authorization: Bearer {OPENROUTER_API_KEY}
     HTTP-Referer: https://github.com/Rawknee-69/Beta-Claw
     X-Title: betaclaw
   Body: {
     "model": "meta-llama/llama-3.1-8b-instruct:free",
     "messages": [{"role": "user", "content": "Say OK"}],
     "max_tokens": 5
   }
   

5. Verify the response is valid.

Step 4: Register Provider

1. Register the OpenRouter adapter in the provider registry.
2. Update .beta/config.toon:

   @providers{
     default:openrouter
     configured:[openrouter]
   }
   

3. Ask the user if OpenRouter should be the default provider (recommended if it's their only provider).

Step 5: Populate Model Catalog

1. Fetch the full model list from OpenRouter.
2. Filter out deprecated and unavailable models.
3. Assign each model to a tier based on capabilities and pricing:
   • Nano (0–20): free models, small Llama/Gemma variants
   • Standard (21–60): Claude Haiku, GPT-4o-mini, Gemini Flash
   • Pro (61–85): Claude Sonnet, GPT-4o, Gemini Pro
   • Max (86–100): Claude Opus, GPT-o3, Gemini Ultra
4. Store the catalog in SQLite with a 4-hour TTL.
5. Report the number of models available per tier.

Step 6: Configure Model Preferences (Optional)

Ask the user if they want to:

• Pin specific models: Always use a particular model for certain tiers.
• Exclude models: Block specific models from being selected.
• Set spending limits: Maximum cost per request or per day.

Store preferences in .beta/config.toon.

Step 7: Confirm

Report:

• OpenRouter is configured with access to N models
• Models per tier: Nano (X), Standard (Y), Pro (Z), Max (W)
• Default provider: OpenRouter (or whatever was set)
• API key stored securely in vault
• Model catalog refreshes automatically every 4 hours
• The agent will automatically select the best model for each task based on complexity
• To add more providers (e.g., direct Anthropic or OpenAI keys for lower latency): run /add-provider
• Account balance/credits can be checked at https://openrouter.ai/account