openai

package
v0.1.6 Latest Latest
Warning

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

 Go to latest Published: Apr 1, 2026 License: Apache-2.0 Imports: 31 Imported by: 0

Documentation

Index

Constants

This section is empty.

Variables

This section is empty.

Functions

func ToLLMSFromResponses 

func ToLLMSFromResponses(resp *ResponsesResponse) *llm.GenerateResponse

ToLLMSFromResponses converts a Responses API response to llm.GenerateResponse.

func ToLLMSResponse 

func ToLLMSResponse(resp *Response) *llm.GenerateResponse

ToLLMSResponse converts a Response to an llm.ChatResponse

Types

type APIKeyProvider 

type APIKeyProvider func(ctx context.Context) (string, error)

type AuthDiagnosticsProvider 

type AuthDiagnosticsProvider func(ctx context.Context) string

type ChatGPTAccountIDProvider 

type ChatGPTAccountIDProvider func(ctx context.Context) (string, error)

type ChatGPTBackendResponsesPayload 

type ChatGPTBackendResponsesPayload struct {
	Model             string          `json:"model"`
	Instructions      string          `json:"instructions"`
	Input             []InputItem     `json:"input"`
	Tools             []ResponsesTool `json:"tools,omitempty"`
	ToolChoice        interface{}     `json:"tool_choice,omitempty"`
	ParallelToolCalls bool            `json:"parallel_tool_calls,omitempty"`
	Reasoning         interface{}     `json:"reasoning,omitempty"`
	Store             bool            `json:"store"`
	Stream            bool            `json:"stream"`
	Include           []string        `json:"include,omitempty"`
	PromptCacheKey    string          `json:"prompt_cache_key,omitempty"`
	Text              *TextControls   `json:"text,omitempty"`
}

ChatGPTBackendResponsesPayload is the dedicated request contract for https://chatgpt.com/backend-api/codex/responses.

Do not add v1/responses-only fields here (e.g. temperature, top_p, n, max_output_tokens). Backend rejects several of those.

func ToChatGPTBackendResponsesPayload 

func ToChatGPTBackendResponsesPayload(req *Request) *ChatGPTBackendResponsesPayload

ToChatGPTBackendResponsesPayload builds a backend-specific payload. It intentionally applies backend constraints and role adaptations that differ from OpenAI v1/responses.

type Choice 

type Choice struct {
	Index        int     `json:"index"`
	Message      Message `json:"message"`
	FinishReason string  `json:"finish_reason"`
}

Choice represents a choice in the OpenAI API response

type Client 

type Client struct {
	basecfg.Config
	APIKey string
	// APIKeyProvider resolves the API key at call time (e.g., from OAuth token exchange).
	// When set, it is used only if APIKey is empty.
	APIKeyProvider APIKeyProvider
	// UserAgent overrides the default User-Agent header when specified and allowed.
	UserAgent string
	// EnableLogging toggles provider runtime logs (auth/ws diagnostics).
	EnableLogging bool
	// Originator mirrors Codex default header style for backend-api compatibility.
	Originator string
	// CodexBetaFeatures maps to x-codex-beta-features header when set.
	CodexBetaFeatures string
	// AuthSource is a redacted label indicating where auth keys are resolved from.
	AuthSource string
	// AuthDiagnosticsProvider returns redacted runtime diagnostics for auth decisions.
	AuthDiagnosticsProvider AuthDiagnosticsProvider
	// ChatGPTAccountIDProvider resolves workspace/account id for ChatGPT backend requests.
	ChatGPTAccountIDProvider ChatGPTAccountIDProvider
	// ChatGPTAccountID is an optional static workspace/account id.
	ChatGPTAccountID string

	// Defaults applied when GenerateRequest.Options is nil or leaves the
	// respective field unset.
	MaxTokens   int
	Temperature *float64

	// ContextContinuation controls whether this client should use
	// response continuation by response_id when supported. When nil,
	// continuation is considered enabled.
	ContextContinuation *bool
	// contains filtered or unexported fields
}

Client represents an OpenAI API client

func NewClient 

func NewClient(apiKey, model string, options ...ClientOption) *Client

NewClient creates a new OpenAI client with the given API key and model

func (*Client) Generate 

func (c *Client) Generate(ctx context.Context, request *llm.GenerateRequest) (*llm.GenerateResponse, error)

Generate sends a chat request to the OpenAI API and returns the response

func (*Client) Implements 

func (c *Client) Implements(feature string) bool

func (*Client) Stream 

func (c *Client) Stream(ctx context.Context, request *llm.GenerateRequest) (<-chan llm.StreamEvent, error)

Stream sends a chat request to the OpenAI API with streaming enabled and returns a channel of partial responses.

func (*Client) SupportsAnchorContinuation 

func (c *Client) SupportsAnchorContinuation() bool

SupportsAnchorContinuation reports whether previous_response_id-style continuation should be attempted for this client.

func (*Client) ToRequest 

func (c *Client) ToRequest(request *llm.GenerateRequest) (*Request, error)

ToRequest converts an llm.ChatRequest to a Request

type ClientOption 

type ClientOption func(*Client)

ClientOption mutates an OpenAI Client instance.

func WithAPIKeyProvider 

func WithAPIKeyProvider(provider APIKeyProvider) ClientOption

WithAPIKeyProvider configures a resolver used to obtain an API key at call time. This is intended for auth flows that mint or refresh API keys dynamically.

func WithAuthDiagnosticsProvider 

func WithAuthDiagnosticsProvider(provider AuthDiagnosticsProvider) ClientOption

WithAuthDiagnosticsProvider sets runtime auth diagnostics producer.

func WithAuthSource 

func WithAuthSource(source string) ClientOption

WithAuthSource sets a redacted label describing auth source selection.

func WithBaseURL 

func WithBaseURL(baseURL string) ClientOption

func WithChatGPTAccountID 

func WithChatGPTAccountID(accountID string) ClientOption

WithChatGPTAccountID sets static ChatGPT workspace/account id header value.

func WithChatGPTAccountIDProvider 

func WithChatGPTAccountIDProvider(provider ChatGPTAccountIDProvider) ClientOption

WithChatGPTAccountIDProvider sets runtime resolver for ChatGPT workspace/account id.

func WithCodexBetaFeatures 

func WithCodexBetaFeatures(features string) ClientOption

WithCodexBetaFeatures sets x-codex-beta-features header value.

func WithContextContinuation 

func WithContextContinuation(enabled *bool) ClientOption

WithContextContinuation sets a client-level toggle for server-side context continuation (continuation by response_id) when supported by the provider.

func WithHTTPClient 

func WithHTTPClient(httpClient *http.Client) ClientOption

func WithLoggingEnabled 

func WithLoggingEnabled(enabled bool) ClientOption

WithLoggingEnabled toggles provider runtime logs.

func WithMaxTokens 

func WithMaxTokens(max int) ClientOption

WithMaxTokens sets a default max_tokens that will be applied to any Generate request that does not explicitly specify MaxTokens in the options.

func WithModel 

func WithModel(model string) ClientOption

func WithOriginator 

func WithOriginator(originator string) ClientOption

WithOriginator sets an explicit originator header value.

func WithTemperature 

func WithTemperature(temp float64) ClientOption

WithTemperature sets a default temperature applied when a Generate request does not specify it.

func WithTimeout 

func WithTimeout(timeoutSeconds int) ClientOption

func WithUsageListener 

func WithUsageListener(l basecfg.UsageListener) ClientOption

WithUsageListener assigns token usage listener to the client.

func WithUserAgent 

func WithUserAgent(userAgent string) ClientOption

WithUserAgent sets a User-Agent override for OpenAI requests. The override is applied only when the value starts with "openai" (case-insensitive).

type Container 

type Container struct {
	Type    string   `json:"type"`
	FileIds []string `json:"file_ids,omitempty"`
}

type ContentItem 

type ContentItem struct {
	Type     string    `json:"type"`
	Text     string    `json:"text,omitempty"`
	ImageURL *ImageURL `json:"image_url,omitempty"`
	File     *File     `json:"file,omitempty"`
}

ContentItem represents a single content item in a message for the OpenAI API

type DeltaMessage 

type DeltaMessage struct {
	Role      string          `json:"role,omitempty"`
	Content   *string         `json:"content,omitempty"`
	ToolCalls []ToolCallDelta `json:"tool_calls,omitempty"`
}

type File 

type File struct {
	FileID   string `json:"file_id,omitempty"`
	FileName string `json:"filename,omitempty"`
	FileData string `json:"file_data,omitempty"`
}

type FunctionCall 

type FunctionCall struct {
	Name      string `json:"name"`
	Arguments string `json:"arguments"`
}

FunctionCall represents a function call in the OpenAI API

type FunctionCallDelta 

type FunctionCallDelta struct {
	Name      string `json:"name,omitempty"`
	Arguments string `json:"arguments,omitempty"`
}

type ImageURL 

type ImageURL struct {
	URL    string `json:"url"`
	Detail string `json:"detail,omitempty"`
}

ImageURL represents an image referenced by URL for the OpenAI API

type InputItem 

type InputItem struct {
	Type string `json:"type"`
	// Message fields.
	Role string `json:"role,omitempty"`
	// Name is not supported by Responses API and must be omitted.
	Content []ResponsesContentItem `json:"content,omitempty"`
	// ToolCallID is required by OpenAI when role == "tool" to associate
	// the tool result with a prior assistant tool_call request.
	ToolCallID string `json:"tool_call_id,omitempty"`
	// function_call_output fields.
	CallID string `json:"call_id,omitempty"`
	Output string `json:"output,omitempty"`
}

type Message 

type Message struct {
	Role         string        `json:"role"`
	Content      interface{}   `json:"content,omitempty"` // Can be string or []ContentItem
	Name         string        `json:"name,omitempty"`
	FunctionCall *FunctionCall `json:"function_call,omitempty"`
	ToolCalls    []ToolCall    `json:"tool_calls,omitempty"`
	ToolCallId   string        `json:"tool_call_id,omitempty"`
}

Message represents a message in the OpenAI API request

type Request 

type Request struct {
	Tools         []Tool         `json:"tools,omitempty"`
	Model         string         `json:"model"`
	Messages      []Message      `json:"messages"`
	Temperature   *float64       `json:"temperature,omitempty"`
	MaxTokens     int            `json:"max_completion_tokens,omitempty"`
	TopP          float64        `json:"top_p,omitempty"`
	N             int            `json:"n,omitempty"`
	Stream        bool           `json:"stream,omitempty"`
	StreamOptions *StreamOptions `json:"stream_options,omitempty"`
	// Reasoning enables configuration of internal chain-of-thought reasoning features.
	Reasoning *llm.Reasoning `json:"reasoning,omitempty"`
	// Instructions provides system guidance for the Responses API.
	Instructions string `json:"instructions,omitempty"`
	// PromptCacheKey enables provider-side prompt caching when supported.
	PromptCacheKey string `json:"prompt_cache_key,omitempty"`
	// Text controls output formatting/verbosity for the Responses API.
	Text *TextControls `json:"text,omitempty"`

	ToolChoice        interface{} `json:"tool_choice,omitempty"`
	ParallelToolCalls bool        `json:"parallel_tool_calls,omitempty"`

	// PreviousResponseID allows continuing a prior Responses API call.
	PreviousResponseID string `json:"previous_response_id,omitempty"`
	// EnableCodeInterpreter controls stream-only injection of a default
	// code_interpreter tool in Responses API payloads.
	EnableCodeInterpreter bool `json:"-"`
	// EnableImageGeneration controls stream-only injection of the built-in
	// image_generation tool in Responses API payloads.
	EnableImageGeneration bool `json:"-"`
}

Request represents the request structure for OpenAI API

func ToRequest 

func ToRequest(request *llm.GenerateRequest) *Request

ToRequest is a convenience wrapper retained for backward-compatible tests. It constructs a default client and adapts an llm.GenerateRequest to provider Request. Errors are ignored in this wrapper; callers requiring error handling should use Client.ToRequest.

type Response 

type Response struct {
	ID      string   `json:"id"`
	Object  string   `json:"object"`
	Created int64    `json:"created"`
	Model   string   `json:"model"`
	Choices []Choice `json:"choices"`
	Usage   Usage    `json:"usage"`
}

Response represents the response structure from OpenAI API

type ResponsesContentItem 

type ResponsesContentItem struct {
	Type     string `json:"type"`
	Text     string `json:"text,omitempty"`
	ImageURL string `json:"image_url,omitempty"`
	FileData string `json:"file_data,omitempty"`
	FileName string `json:"filename,omitempty"`
	Detail   string `json:"detail,omitempty"`
	FileID   string `json:"file_id,omitempty"`
	// Function call output back to the model
	CallID string `json:"call_id,omitempty"`
	Output string `json:"output,omitempty"`
	// Allow passthrough for future shapes (e.g., input_audio)
	Extra map[string]interface{} `json:"-"`
}

type ResponsesImageURL 

type ResponsesImageURL struct {
	URL    string `json:"url"`
	Detail string `json:"detail,omitempty"`
}

type ResponsesOutputItem 

type ResponsesOutputItem struct {
	// Type is typically "message".
	Type    string                 `json:"type"`
	Role    string                 `json:"role,omitempty"`
	Content []ResponsesContentItem `json:"content,omitempty"`
	// Tool calls when assistant requests tools
	ToolCalls []ToolCall `json:"tool_calls,omitempty"`
	// Function-call item shape (when Type == "function_call")
	CallID    string `json:"call_id,omitempty"`
	Name      string `json:"name,omitempty"`
	Status    string `json:"status,omitempty"`
	Arguments string `json:"arguments,omitempty"`
}

type ResponsesPayload 

type ResponsesPayload struct {
	Model              string          `json:"model"`
	Instructions       string          `json:"instructions,omitempty"`
	Input              []InputItem     `json:"input"`
	Tools              []ResponsesTool `json:"tools,omitempty"`
	Temperature        *float64        `json:"temperature,omitempty"`
	MaxOutputTokens    int             `json:"max_output_tokens,omitempty"`
	TopP               float64         `json:"top_p,omitempty"`
	N                  int             `json:"n,omitempty"`
	Stream             bool            `json:"stream,omitempty"`
	ToolChoice         interface{}     `json:"tool_choice,omitempty"`
	ParallelToolCalls  bool            `json:"parallel_tool_calls,omitempty"`
	Reasoning          *llm.Reasoning  `json:"reasoning,omitempty"`
	PreviousResponseID string          `json:"previous_response_id,omitempty"`
	Store              *bool           `json:"store,omitempty"`
	Include            []string        `json:"include,omitempty"`
	PromptCacheKey     string          `json:"prompt_cache_key,omitempty"`
	Text               *TextControls   `json:"text,omitempty"`
	// Provider-specific metadata passthrough if needed in future
	Extra map[string]interface{} `json:"-"`
}

ResponsesPayload is the request body for the /v1/responses API.

func ToResponsesPayload 

func ToResponsesPayload(req *Request) *ResponsesPayload

ToResponsesPayload converts legacy Request into Responses API payload.

type ResponsesResponse 

type ResponsesResponse struct {
	ID     string                `json:"id"`
	Status string                `json:"status,omitempty"`
	Model  string                `json:"model"`
	Output []ResponsesOutputItem `json:"output"`
	Usage  ResponsesUsage        `json:"usage"`
}

ResponsesResponse represents a non-streaming reply from the /v1/responses API or the final object emitted in the streaming completed event.

type ResponsesTool 

type ResponsesTool struct {
	Type         string                 `json:"type"`
	Name         string                 `json:"name,omitempty"`
	Description  string                 `json:"description,omitempty"`
	Parameters   map[string]interface{} `json:"parameters,omitempty"`
	Required     []string               `json:"required,omitempty"`
	Strict       *bool                  `json:"strict,omitempty"`
	Container    *Container             `json:"container,omitempty"`
	OutputFormat string                 `json:"output_format,omitempty"`
	Background   string                 `json:"background,omitempty"`
	Size         string                 `json:"size,omitempty"`
}

ResponsesTool is the Responses API tool schema. For function tools, the name/description/parameters are top-level (not nested under "function").

type ResponsesUsage 

type ResponsesUsage struct {
	InputTokens  int `json:"input_tokens"`
	OutputTokens int `json:"output_tokens"`
	TotalTokens  int `json:"total_tokens"`
}

type StreamChoice 

type StreamChoice struct {
	Index        int          `json:"index"`
	Delta        DeltaMessage `json:"delta"`
	FinishReason *string      `json:"finish_reason"`
}

type StreamOptions 

type StreamOptions struct {
	IncludeUsage bool `json:"include_usage,omitempty"`
}

StreamOptions controls additional streaming behavior.

type StreamResponse 

type StreamResponse struct {
	ID      string         `json:"id"`
	Object  string         `json:"object"`
	Created int64          `json:"created"`
	Model   string         `json:"model"`
	Choices []StreamChoice `json:"choices"`
}

StreamResponse represents a single Server-Sent Event chunk from OpenAI chat/completions endpoint when stream=true. The payload places partial deltas under choices[i].delta instead of choices[i].message.

type TextControls 

type TextControls struct {
	Verbosity string      `json:"verbosity,omitempty"`
	Format    *TextFormat `json:"format,omitempty"`
}

TextControls enables response formatting controls on the Responses API.

type TextFormat 

type TextFormat struct {
	Type   string                 `json:"type"`
	Strict bool                   `json:"strict"`
	Schema map[string]interface{} `json:"schema"`
	Name   string                 `json:"name,omitempty"`
}

TextFormat configures structured text output on the Responses API.

type Tool 

type Tool struct {
	Type     string         `json:"type"`
	Function ToolDefinition `json:"function"`
}

Tool represents a tool in the OpenAI API

type ToolCall 

type ToolCall struct {
	ID       string       `json:"id"`
	Type     string       `json:"type"`
	Function FunctionCall `json:"function"`
}

ToolCall represents a tool call in the OpenAI API

type ToolCallDelta 

type ToolCallDelta struct {
	Index    int               `json:"index"`
	ID       string            `json:"id,omitempty"`
	Type     string            `json:"type,omitempty"`
	Function FunctionCallDelta `json:"function,omitempty"`
}

ToolCallDelta mirrors the incremental tool call fields included in streaming deltas. Arguments are delivered as a concatenated string across multiple events.

type ToolDefinition 

type ToolDefinition struct {
	Name        string                 `json:"name"`
	Description string                 `json:"description,omitempty"`
	Parameters  map[string]interface{} `json:"parameters,omitempty"`
	Required    []string               `json:"required,omitempty"`
	Strict      bool                   `json:"strict,omitempty"`
}

ToolDefinition represents a tool definition in the OpenAI API

type Usage 

type Usage struct {
	PromptTokens     int `json:"prompt_tokens"`
	CompletionTokens int `json:"completion_tokens"`
	TotalTokens      int `json:"total_tokens"`
	// Some OpenAI responses provide flattened fields in addition to details.
	// Support both shapes to ensure robust parsing across models/endpoints.
	PromptCachedTokens        int `json:"prompt_cached_tokens,omitempty"`
	ReasoningTokens           int `json:"reasoning_tokens,omitempty"`
	CompletionReasoningTokens int `json:"completion_reasoning_tokens,omitempty"`
	PromptTokensDetails       struct {
		CachedTokens int `json:"cached_tokens"`
		AudioTokens  int `json:"audio_tokens"`
	} `json:"prompt_tokens_details"`
	CompletionTokensDetails struct {
		ReasoningTokens          int `json:"reasoning_tokens"`
		AudioTokens              int `json:"audio_tokens"`
		AcceptedPredictionTokens int `json:"accepted_prediction_tokens"`
		RejectedPredictionTokens int `json:"rejected_prediction_tokens"`
	} `json:"completion_tokens_details"`
}

Usage represents token usage information in the OpenAI API response

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.