autobot-provider

name	autobot-provider
description	Add new LLM providers to Autobot with proper integration
tags	["provider","llm","integration"]
metadata	{"scope":"feature-development"}

Adding a New LLM Provider

Files to Create/Modify

1. Create provider implementation:

   • src/autobot/providers/<name>_provider.cr
   • Inherit from HttpProvider or implement Provider interface

2. Register provider:

   • Add a new ProviderSpec to the PROVIDERS list in src/autobot/providers/registry.cr.
   • Ensure keywords and API URL are correctly specified.

3. Add configuration schema:

   • Update src/autobot/config/schema.cr with provider config if needed.
   • Most HTTP providers work automatically if they follow OpenAI compatibility.

4. Create tests:

   • spec/autobot/providers/<name>_provider_spec.cr
   • Test provider registration in registry_spec.cr.
   • Test API integration (mocked).

5. Add documentation:

   • docs/<name>.md - Provider setup guide.
   • Update docs/providers.md - Add to provider comparison.

Provider Implementation Template

# src/autobot/providers/<name>_provider.cr
module Autobot
  module Providers
    class <Name>Provider < HttpProvider
      def initialize(
        api_key : String,
        api_base : String? = nil,
        @model : String = "default-model",
        @extra_headers = {} of String => String,
        provider_name : String? = nil,
      )
        super(api_key, api_base)
        @gateway = Providers.find_gateway(provider_name, api_key, api_base)
      end

      def chat(
        messages : Array(Hash(String, JSON::Any)),
        tools : Array(Hash(String, JSON::Any))? = nil,
        model : String? = nil,
        max_tokens : Int32 = DEFAULT_MAX_TOKENS,
        temperature : Float64 = DEFAULT_TEMPERATURE,
      ) : Response
        # Implement API-specific logic here
        # Or call super for OpenAI-compatible APIs
        super
      end

      def default_model : String
        @model
      end
    end
  end
end

Alternative: Non-HTTP Provider

For providers with non-standard APIs (like AWS Bedrock), inherit directly from Provider:

class BedrockProvider < Provider
  # Implement abstract methods directly
  # See src/autobot/providers/bedrock_provider.cr for full example
end

ProviderSpec Registration

Add to PROVIDERS in src/autobot/providers/registry.cr:

ProviderSpec.new(
  name: "<name>",
  keywords: ["<keyword1>", "<keyword2>"],
  display_name: "<Display Name>",
  api_url: "https://api.<provider>.com/v1/chat/completions",
)

Testing Checklist

• Provider registers correctly in registry
• Configuration schema validates properly
• API calls are mocked in tests
• Error handling works (401, 429, 500 responses)
• User-Agent handling (if provider requires specific UA)
• Interactive setup includes new provider

Documentation Requirements

Each provider doc must include:

• API key acquisition instructions
• Supported models list
• Rate limits and pricing notes
• Special features (if any)
• Example configuration

When to Use

Use this skill when:

• Adding support for a new LLM service
• Integrating a new AI provider API
• Testing provider implementations
• Updating provider documentation

Related Skills: autobot-test