ReqLLM.
View Source
Azure Responses API adapter.
Thin wrapper around ReqLLM.Providers.OpenAI.ResponsesAPI that delegates all
encoding/decoding to the native OpenAI implementation.
Endpoint
Azure Responses API uses: {base_url}/responses?api-version=...
where base_url ends with /openai (model specified in request body, not in URL path)
Supported Models
Models with "api": "responses" in their metadata:
- codex-mini, gpt-5-codex, gpt-5.1-codex-mini
- Future models that use the Responses API format
Reasoning Effort
Accepts the top-level :reasoning_effort option (:minimal | :low | :medium | :high | :xhigh | :none or the equivalent string) and forwards it to the
Responses API as "reasoning": {"effort": <level>}.
Parallel Tool Calls
Accepts parallel_tool_calls via the top-level option, or via
provider_options[:openai_parallel_tool_calls] (Azure's preferred key) or
provider_options[:parallel_tool_calls]. Forwarded to the Responses API
as the top-level "parallel_tool_calls" boolean. An explicit false is
preserved.
Summary
Functions
Decodes Server-Sent Events for streaming responses.
Extracts usage from Azure Responses API payloads.
Azure Responses API models do not support embeddings.
Formats a request body for the Azure Responses API.
Parses a Responses API response body.
Functions
Decodes Server-Sent Events for streaming responses.
Delegates to the native OpenAI ResponsesAPI.
Extracts usage from Azure Responses API payloads.
Azure Responses API models do not support embeddings.
Formats a request body for the Azure Responses API.
Delegates to the native OpenAI Responses API encoder.
Parses a Responses API response body.
Delegates to the native OpenAI ResponsesAPI decoder.