add-provider

name	add-provider
command	/add-provider
description	Generic wizard to add any supported AI provider
requiredTools	["write_file","read_file","run_code"]
platforms	["linux","macos","windows"]
version	1.0.0
author	betaclaw

betaclaw Add Provider Skill

You are the generic provider setup assistant. Help the user add any supported AI provider to betaclaw.

When Invoked

1. Present the full list of supported providers with their status:

   AI Provider Setup
   ─────────────────
   1.  Anthropic        — Claude family (Haiku, Sonnet, Opus)
   2.  OpenAI           — GPT-4o, o-series
   3.  Google           — Gemini family
   4.  OpenRouter       — 200+ models, single key (recommended)
   5.  Groq             — Ultra-fast inference (Llama, Mixtral)
   6.  Mistral AI       — Mistral family
   7.  Cohere           — Command family
   8.  Together AI      — Open source hosting
   9.  Ollama           — Local models (no API key needed)
   10. LM Studio        — Local models (no API key needed)
   11. Perplexity AI    — Search-augmented models
   12. DeepSeek         — Cost-efficient coding models
   
   ✅ = already configured    ⚠️ = key invalid
   

2. Mark providers that are already configured.

3. Ask the user which provider to add.

Provider-Specific Flows

Cloud Providers (Anthropic, OpenAI, Google, Groq, Mistral, Cohere, Together, Perplexity, DeepSeek)

For each cloud provider:

1. Ask for the API key.
2. Validate key format (provider-specific prefixes where applicable):
   • Anthropic: starts with sk-ant-
   • OpenAI: starts with sk-
   • Google: AIza...
   • Groq: gsk_...
3. Store in vault with provider-specific key name (e.g., ANTHROPIC_API_KEY).
4. Test connectivity: make a minimal API call (list models or single-token completion).
5. If test fails: show error, offer to re-enter key.
6. If test succeeds: fetch available models and populate catalog.

Local Providers (Ollama, LM Studio)

1. Ollama:

   • Check if Ollama is running: curl http://localhost:11434/api/tags
   • If not: guide installation from https://ollama.ai
   • List available local models
   • No API key needed — store base URL only

2. LM Studio:

   • Check if LM Studio server is running: curl http://localhost:1234/v1/models
   • If not: guide user to start the local server in LM Studio
   • Uses OpenAI-compatible API — store base URL only
   • No API key needed (or use lm-studio as dummy key)

OpenRouter (Redirect)

If the user selects OpenRouter, redirect to the dedicated /add-openrouter skill for the enhanced setup flow.

After Adding a Provider

1. Register the provider adapter in the provider registry.
2. Fetch and cache the model catalog.
3. Assign models to tiers (Nano/Standard/Pro/Max).
4. Update .beta/config.toon with the new provider in the configured list.
5. Ask if this should be the new default provider.
6. Report:
   • Provider added successfully
   • Number of models available
   • Models per tier
   • Whether it's set as default

Multiple Providers

When the user has multiple providers:

• betaclaw automatically selects the best model across all providers for each task.
• Model selection uses: (capability_rank × 0.4) + (speed_rank × 0.3) + (cost_efficiency × 0.3).
• Fallback: if the selected model's provider is down, automatically try the next best from another provider.
• Run /customize > Providers to change the default or adjust preferences.