openairesponsessdk

package
v0.25.1 Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Sep 7, 2026 License: MIT Imports: 15 Imported by: 0

README

OpenAI Responses Adapter

Use this adapter for OpenAI Responses and providers that expose a compatible Responses-style API.

Preset-backed providers using this adapter include:

  • OpenAI Responses
  • DeepSeek
  • Meta
  • MiniMax
  • OpenRouter
  • QwenCloud
  • Xiaomi
  • Z.AI Coding Plan
  • xAI
  • LocalAI
  • LM Studio
  • SGLang
  • vLLM

Best fit

Choose this adapter when your application benefits from:

  • OpenAI Responses-style reasoning
  • Encrypted reasoning continuation
  • Rich image and file input
  • JSON Schema output
  • Built-in web search where supported by the provider/model
  • Function and custom tools
  • Thinking streaming where exposed by the upstream response
  • Top-level prompt-cache controls

High-level support

Capability Support
Text input and output Supported
Image input Supported
File input Supported
Text streaming Supported
Thinking streaming Supported where upstream reasoning deltas are emitted
Reasoning configuration Supported for compatible models
Reasoning history Encrypted reasoning only
JSON Schema output Supported
Output verbosity Supported where the effective model allows it
Stop sequences Not supported by the base adapter
Function tools Supported
Custom tools Supported through compatible function-style transport
Web search Supported where the provider/model exposes it
Tool policy Supported
Cache control Top-level prompt-cache behavior only
URL citations Supported where returned by the provider
Audio/video/image output Not normalized

Important guidance

Reasoning continuation

OpenAI Responses continuation is safest when the prior response contains encrypted reasoning content.

The adapter keeps encrypted reasoning and removes incompatible visible reasoning data before replaying history.

Do not expect Anthropic signed thinking or Google thought signatures to continue through this adapter.

Reasoning summary, context, and mode

The official OpenAI Responses preset supports all normalized reasoning controls:

  • Summary styles:
    • omitted maps to OpenAI summary="concise".
    • auto, concise, and detailed retain their OpenAI mappings.
    • An unspecified summary style uses OpenAI summary="auto".
  • Context:
    • auto
    • current_turn
    • all_turns
  • Mode:
    • standard
    • pro

Context and mode are OpenAI Responses-specific controls. Responses-compatible provider presets disable them by default unless their endpoint/model support has been explicitly verified through a capability override.

Stop sequences

The base Responses API does not support normalized stop sequences.

If your application depends on stop sequences, inspect warnings and consider a Chat Completions-compatible provider/model instead.

Cache controls

This adapter supports top-level prompt-cache behavior.

Per-message, tool, and reasoning cache controls are not portable through the Responses adapter.

Tools

Function and custom tools are supported.

Custom tools may use function-style transport underneath because that is the compatible request shape available through the adapter.

Web search is provider-native. Its exact returned detail and policy behavior can vary by provider/model.

Stateful provider features

The normalized API is intentionally stateless.

Do not depend on provider-native features such as stored responses, previous-response IDs, background jobs, or prompt resources through this adapter.

Working examples

Documentation

Index

Constants

This section is empty.

Variables

This section is empty.

Functions

This section is empty.

Types

type OpenAIResponsesAPI 

type OpenAIResponsesAPI struct {
	ProviderParam *spec.ProviderParam
	// contains filtered or unexported fields
}

OpenAIResponsesAPI struct that implements the CompletionProvider interface.

func NewOpenAIResponsesAPI 

func NewOpenAIResponsesAPI(
	pi spec.ProviderParam,
	debugger spec.CompletionDebugger,
) (*OpenAIResponsesAPI, error)

func (*OpenAIResponsesAPI) DeInitLLM 

func (api *OpenAIResponsesAPI) DeInitLLM(ctx context.Context) error

func (*OpenAIResponsesAPI) FetchCompletion 

func (api *OpenAIResponsesAPI) FetchCompletion(
	ctx context.Context,
	inReq *spec.FetchCompletionRequest,
	opts *spec.FetchCompletionOptions,
) (*spec.FetchCompletionResponse, error)

func (*OpenAIResponsesAPI) GetProviderCapability added in v0.9.1 

func (api *OpenAIResponsesAPI) GetProviderCapability(ctx context.Context) (spec.ModelCapabilities, error)

func (*OpenAIResponsesAPI) GetProviderInfo 

func (api *OpenAIResponsesAPI) GetProviderInfo(ctx context.Context) *spec.ProviderParam

func (*OpenAIResponsesAPI) InitLLM 

func (api *OpenAIResponsesAPI) InitLLM(ctx context.Context) error

func (*OpenAIResponsesAPI) IsConfigured 

func (api *OpenAIResponsesAPI) IsConfigured(ctx context.Context) bool

func (*OpenAIResponsesAPI) SetProviderAPIKey 

func (api *OpenAIResponsesAPI) SetProviderAPIKey(
	ctx context.Context,
	apiKey string,
) error

SetProviderAPIKey sets the key for a provider.

Source Files

View all Source files

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.