Protocol for encoding canonical ReqLLM.Context structures to provider-specific request JSON.

This protocol handles the request encoding phase, converting ReqLLM contexts and models into the JSON format expected by each provider's API.

Default Implementation

The Map implementation provides a baseline OpenAI-compatible request format that works for most providers including OpenAI, Groq, OpenRouter, and xAI:

ReqLLM.Context.Codec.encode_request(context, model)
#=> %{
#     model: "gpt-4",
#     messages: [%{role: "user", content: "Hello"}],
#     stream: true,
#     max_tokens: 1000,
#     temperature: 0.7,
#     tools: [%{type: "function", function: %{name: "...", ...}}]
#   }

Provider-Specific Overrides

Providers that require different formats can implement their own protocol:

defimpl ReqLLM.Context.Codec, for: MyProvider.Context do
  def encode_request(context, model) do
    # Custom encoding logic for provider-specific format
  end
end

Tool Encoding

Tools are automatically converted to OpenAI function format using ReqLLM.Schema.to_openai_format/1, which handles parameter schema conversion from keyword lists to JSON Schema.