airouter

package module

v0.0.65 Latest Latest Go to latest Published: Apr 1, 2026 License: MIT Imports: 21 Imported by: 0

Details

Valid go.mod file
Redistributable license
Tagged version
Stable version
Learn more about best practices

Repository

github.com/subiz/airouter

Links

Open Source Insights

README ¶

AI Router

AI Router is a lightweight Go client for the Subiz AI Proxy, providing a unified interface to multiple LLM providers including OpenAI, Google Gemini, and more.

Features

Unified API: Single interface for Chat Completion, Text Embedding, and Reranking across different providers.
Provider Agnostic: Switch between GPT-4, GPT-5, Gemini, and other models without changing your core logic.
Simplified Configuration: Easy initialization with a Subiz API key.

Installation

go get github.com/subiz/airouter

Getting Started

Initialization

Initialize the library with your Subiz API key:

import "github.com/subiz/airouter"

func init() {
    airouter.Init("YOUR_SUBIZ_API_KEY")
}

Usage Examples

1. Chat Completion

package main

import (
    "context"
    "fmt"
    "github.com/subiz/airouter"
)

func main() {
    airouter.Init("YOUR_SUBIZ_API_KEY")

    ctx := context.Background()
    output, _, err := airouter.Complete(ctx, airouter.CompletionInput{
        Model: airouter.Gpt_5_nano,
        ReasoningEffort: "low",
        Instruct: "Tell a short story about a brave dragon.",
    })

    if err != nil {
        panic(err)
    }

    fmt.Println(output) // The dragon’s name was Ember, and his light was not a blaze but a memory of dawn...
}

2. Text Embedding

package main

import (
    "context"
    "fmt"
    "github.com/subiz/airouter"
)

func main() {
    airouter.Init("YOUR_SUBIZ_API_KEY")

    vector, _, err := airouter.GetEmbedding(context.Background(), airouter.Text_embedding_3_small, "Hello world")
    if err != nil {
        panic(err)
    }

    fmt.Printf("Vector length: %d\n", len(vector))
    fmt.Println(vector[:5]) // Print first 5 dimensions
}

3. Rerank

package main

import (
    "context"
    "fmt"
    "github.com/subiz/airouter"
)

func main() {
    airouter.Init("YOUR_SUBIZ_API_KEY")

    records := []*airouter.RerankRecord{
        {Id: "1", Title: "Greeting", Content: "Xin chào thế giới"},
        {Id: "2", Title: "Farewell", Content: "Goodbye world"},
    }

    output, err := airouter.Rerank(context.Background(), "google", "hello", records)
    if err != nil {
        panic(err)
    }

    fmt.Println("Rerank results:", output)
}

Supported Models

Chat Completion

Provider	Model Aliases
OpenAI	`gpt-4o`, `gpt-4o-mini`, `gpt-4.1`, `gpt-4.1-mini`, `gpt-4.1-nano`, `gpt-5`, `gpt-5-mini`, `gpt-5-nano`
Google	`gemini-2.5-pro`, `gemini-2.5-flash`, `gemini-2.5-flash-lite`, `gemini-3.1-flash-lite-preview`

Text Embedding

gemini-embedding-001
text-embedding-3-small
text-embedding-3-large
text-embedding-ada-002

Reranking

google (Powered by Google Vertex AI)

License

This project is licensed under the MIT License - see the LICENSE file for details.

Documentation ¶

Index ¶

Constants
Variables
func CalculateCost(model string, usage *Usage, service_tier string) int64
func CleanString(str string) string
func FakeBackend(rawURL string, input []byte) (int, []byte)
func GetFallbackChatModel(model string) string
func GetMD5Hash(text string) string
func Init(apikey string)
func InitAPI(geminiKey, openaiKey string)
func ToGeminiModel(model string) string
func ToGeminiRequestJSON(req CompletionInput) ([]byte, error)
func ToModel(model string) string
func ToOpenAICompletionJSON(req CompletionInput) ([]byte, error)
func ToOpenAIModel(model string) string
type CompletionInput
type CompletionOutput
- func Complete(ctx context.Context, input CompletionInput) (string, CompletionOutput, error)
type CompletionReasoning
type Embedding
type EmbeddingOutput
- func GetEmbedding(ctx context.Context, model string, text string) ([]float32, EmbeddingOutput, error)
type Function
type GeminiAPIResponse
type GeminiContent
type GeminiEmbeddingContent
type GeminiEmbeddingPart
type GeminiEmbeddingRequest
type GeminiEmbeddingResponse
type GeminiError
type GeminiErrorDetail
type GeminiFunctionCall
type GeminiFunctionDeclaration
type GeminiFunctionResponse
type GeminiGenerationConfig
type GeminiPart
type GeminiPartInlineData
type GeminiRankingResponse
type GeminiRequest
type GeminiRerankRecord
type GeminiThinkingConfig
type GeminiTool
type LocalCandidate
type LocalContent
type LocalPart
type LocalUsageMetadata
type OpenAIChatMessage
- func (m *OpenAIChatMessage) GetContent() string
type OpenAIChatResponse
- func ChatCompleteAPI(ctx context.Context, payload []byte) (OpenAIChatResponse, error)
type OpenAIChoice
type OpenAIEmbeddingData
type OpenAIEmbeddingRequest
type OpenAIEmbeddingResponse
- func GetEmbeddingAPI(ctx context.Context, model, text string) (OpenAIEmbeddingResponse, error)
type OpenAIError
type OpenAIJSONSchema
type OpenAIMessageContent
type OpenAIMessageContentImageUrl
type OpenAITool
type PromptTokensDetails
type RerankInput
type RerankOutput
- func Rerank(ctx context.Context, model, query string, inrecords []*RerankRecord) (RerankOutput, error)
type RerankRecord
type RerankingResponse
- func RerankAPI(ctx context.Context, token string, payload []byte) (RerankingResponse, error)
type ResponseFormat
type ToolCall
type ToolFunction
type TotalCost
type Usage

Constants ¶

View Source

const Gemini_2_5_flash = "gemini-2.5-flash"

View Source

const Gemini_2_5_flash_lite = "gemini-2.5-flash-lite"

View Source

const Gemini_2_5_pro = "gemini-2.5-pro"

View Source

const Gemini_3_1_flash_lite = "gemini-3.1-flash-lite-preview"

View Source

const Gemini_embedding_001 = "gemini-embedding-001"

View Source

const Gpt_4_1 = "gpt-4.1"

View Source

const Gpt_4_1_mini = "gpt-4.1-mini"

View Source

const Gpt_4_1_nano = "gpt-4.1-nano"

View Source

const Gpt_4o = "gpt-4o"

View Source

const Gpt_4o_mini = "gpt-4o-mini"

View Source

const Gpt_5 = "gpt-5"

View Source

const Gpt_5_4_mini = "gpt-5.4-mini"

View Source

const Gpt_5_4_nano = "gpt-5.4-nano"

View Source

const Gpt_5_mini = "gpt-5-mini"

View Source

const Gpt_5_nano = "gpt-5-nano"

View Source

const O3 = "o3"

View Source

const O4_mini = "o4-mini"

View Source

const Text_embedding_3_large = "text-embedding-3-large"

View Source

const Text_embedding_3_small = "text-embedding-3-small"

View Source

const Text_embedding_ada_002 = "text-embedding-ada-002"

Variables ¶

View Source

var BACKEND = "https://api.subiz.com.vn/4.1/ai"

Functions ¶

func CleanString ¶

func CleanString(str string) string

badstring: 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

func FakeBackend ¶

func FakeBackend(rawURL string, input []byte) (int, []byte)

for testing only

func GetFallbackChatModel ¶

func GetFallbackChatModel(model string) string

GetFallbackChatModel returns a fallback model for a given model. If the model is a GPT model, it returns the equivalent Gemini model. If the model is a Gemini model, it returns the equivalent OpenAI model. Otherwise, it returns a default Gemini model.

func GetMD5Hash ¶

func GetMD5Hash(text string) string

func Init ¶

func Init(apikey string)

Init setups the Subiz API Key client. Only client should call this

func InitAPI ¶

func InitAPI(geminiKey, openaiKey string)

InitAPI setups API Keys for the server. Only server should call this

func ToGeminiModel ¶

func ToGeminiModel(model string) string

ToGeminiModel converts a model name to its equivalent Gemini model.

func ToGeminiRequestJSON ¶

func ToGeminiRequestJSON(req CompletionInput) ([]byte, error)

ToGeminiRequestJSON converts an OpenAIChatRequest to a Gemini-compatible JSON request string.

func ToModel ¶

func ToModel(model string) string

ToModel converts model to the closest standardized model

func ToOpenAICompletionJSON ¶

func ToOpenAICompletionJSON(req CompletionInput) ([]byte, error)

func ToOpenAIModel ¶

func ToOpenAIModel(model string) string

ToOpenAIModel converts a model name to its equivalent OpenAI model.

Types ¶

type CompletionInput ¶

type CompletionInput struct {
	Seed                 int                           `json:"seed,omitempty"`
	PromptCacheKey       string                        `json:"prompt_cache_key,omitempty"`
	PromptCacheRetention string                        `json:"prompt_cache_retention,omitempty"`
	Verbosity            string                        `json:"verbosity,omitempty"`
	Stop                 []string                      `json:"stop,omitempty"`
	Model                string                        `json:"model,omitempty"`
	NoLog                bool                          `json:"-"` // disable log
	Instruct             string                        `json:"instruct,omitempty"`
	Messages             []*header.LLMChatHistoryEntry `json:"messages,omitempty"`
	MaxCompletionTokens  int                           `json:"max_completion_tokens,omitempty"`
	ResponseFormat       *ResponseFormat               `json:"response_format,omitempty"`
	ToolChoice           string                        `json:"tool_choice,omitempty"`
	Reasoning            *CompletionReasoning          `json:"reasoning,omitempty"`
	ReasoningEffort      string                        `json:"reasoning_effort,omitempty"`
	Temperature          float32                       `json:"temperature,omitempty"`
	TopP                 float32                       `json:"top_p,omitempty"`
	Tools                []OpenAITool                  `json:"tools,omitempty"`
	ServiceTier          string                        `json:"service_tier,omitempty"` // [auto, default], flex, priority, scale
	StopAfterToolCalled  bool                          `json:"stop_after_tool_called"`
}

type CompletionOutput ¶

type CompletionOutput struct {
	Content           string `json:"content"`
	Refusal           string `json:"refusal"`
	Request           []byte `json:"request"`
	InputTokens       int64  `json:"input_tokens"`
	OutputTokens      int64  `json:"output_tokens"`
	InputCachedTokens int64  `json:"input_cached_tokens"`
	OuputCachedTokens int64  `json:"output_cached_tokens"`
	Created           int64  `json:"created"`
	DurationMs        int64  `json:"duration_ms"`
	KfpvCostUSD       int64  `json:"kfpv_cost_usd"` // 1 usd -> 1000_000_000 kfpvusd
}

func Complete ¶

func Complete(ctx context.Context, input CompletionInput) (string, CompletionOutput, error)

type CompletionReasoning ¶

type CompletionReasoning struct {
	Effort    string `json:"effort,omitempty"` // high, medium, low
	MaxTokens int    `json:"max_tokens,omitempty"`
}

type Embedding ¶

type Embedding struct {
	Values []float32 `json:"values"`
}

Embedding represents the embedding values.

type EmbeddingOutput ¶

type EmbeddingOutput struct {
	Text        string    `json:"text"`
	Vector      []float32 `json:"vector"`
	TotalTokens int64     `json:"total_tokens"`
	Created     int64     `json:"created"`
	DurationMs  int64     `json:"duration_ms"`
	KfpvCostUSD int64     `json:"kfpv_cost_usd"`
}

func GetEmbedding ¶

func GetEmbedding(ctx context.Context, model string, text string) ([]float32, EmbeddingOutput, error)

type Function ¶

type Function struct {
	Name        string             `json:"name,omitempty"`
	Description string             `json:"description,omitempty"`
	Parameters  *header.JSONSchema `json:"parameters,omitempty"`

	Handler func(ctx context.Context, arg, callid string, ctxm map[string]any) string `json:"-"` // "", true -> abandon completion
}

Function mimics the structure of a function in an OpenAI Chat Completion request.

type GeminiAPIResponse ¶

type GeminiAPIResponse struct {
	Candidates    []*LocalCandidate   `json:"candidates"`
	UsageMetadata *LocalUsageMetadata `json:"usageMetadata"`
	ModelVersion  string              `json:"modelVersion"`
	ResponseId    string              `json:"responseId"`
	Error         *GeminiError        `json:"error"`
}

GeminiAPIResponse is a wrapper for the Gemini API response.

type GeminiContent ¶

type GeminiContent struct {
	Parts []*GeminiPart `json:"parts"`
	Role  string        `json:"role,omitempty"`
}

GeminiContent represents a content block in a Gemini request.

type GeminiEmbeddingContent ¶

type GeminiEmbeddingContent struct {
	Parts []GeminiEmbeddingPart `json:"parts"`
}

GeminiEmbeddingContent represents the content for an embedding request.

type GeminiEmbeddingPart ¶

type GeminiEmbeddingPart struct {
	Text string `json:"text"`
}

GeminiEmbeddingPart represents a part of the content for an embedding request.

type GeminiEmbeddingRequest ¶

type GeminiEmbeddingRequest struct {
	Model   string                 `json:"model"`
	Content GeminiEmbeddingContent `json:"content"`
}

EmbeddingRequest represents the request to the embedding model.

type GeminiEmbeddingResponse ¶

type GeminiEmbeddingResponse struct {
	Embedding *Embedding `json:"embedding"`
}

GeminiEmbeddingResponse represents the response from the embedding model.

type GeminiError ¶

type GeminiError struct {
	Code    int                 `json:"code"`
	Message string              `json:"message"`
	Status  string              `json:"status"`
	Details []GeminiErrorDetail `json:"details"`
}

GeminiError represents a Gemini API error.

type GeminiErrorDetail ¶

type GeminiErrorDetail struct {
	Type            string `json:"@type"`
	FieldViolations []struct {
		Field       string `json:"field"`
		Description string `json:"description"`
	} `json:"fieldViolations"`
}

GeminiErrorDetail represents the details of a Gemini API error.

type GeminiFunctionCall ¶

type GeminiFunctionCall struct {
	Name string         `json:"name"`
	Args map[string]any `json:"args"`
}

GeminiFunctionCall is a local struct to avoid genai dependency.

type GeminiFunctionDeclaration ¶

type GeminiFunctionDeclaration struct {
	Name        string             `json:"name"`
	Description string             `json:"description"`
	Parameters  *header.JSONSchema `json:"parameters"`
}

GeminiFunctionDeclaration is a local struct to avoid genai dependency.

type GeminiFunctionResponse ¶

type GeminiFunctionResponse struct {
	Name     string         `json:"name"`
	Response map[string]any `json:"response"`
}

GeminiFunctionResponse is a local struct to avoid genai dependency.

type GeminiGenerationConfig ¶

type GeminiGenerationConfig struct {
	StopSequences []string `json:"stopSequences,omitempty"`
	// CandidateCount   int                   `json:"candidateCount,omitempty"`
	MaxOutputTokens  int                   `json:"maxOutputTokens,omitempty"`
	Temperature      float32               `json:"temperature,omitempty"`
	TopP             float32               `json:"topP,omitempty"`
	Seed             int                   `json:"seed,omitempty"`
	ResponseMIMEType string                `json:"responseMimeType,omitempty"`
	ResponseSchema   *header.JSONSchema    `json:"responseSchema,omitempty"`
	ThinkingConfig   *GeminiThinkingConfig `json:"thinkingConfig,omitempty"`
}

GeminiGenerationConfig represents the generation configuration.

type GeminiPart ¶

type GeminiPart struct {
	Text             *string                 `json:"text,omitempty"`
	FunctionCall     *GeminiFunctionCall     `json:"functionCall,omitempty"`
	FunctionResponse *GeminiFunctionResponse `json:"functionResponse,omitempty"`
	InlineData       *GeminiPartInlineData   `json:"inline_data,omitempty"`
}

GeminiPart represents a part of a content block.

type GeminiPartInlineData ¶

type GeminiPartInlineData struct {
	MimeType string `json:"mime_type,omitempty"`
	Data     string `json:"data,omitempty"`
}

type GeminiRankingResponse ¶

type GeminiRankingResponse struct {
	Records []*GeminiRerankRecord `json:"records,omitempty"`
	Error   *GeminiError          `json:"error"`
}

type GeminiRequest ¶

type GeminiRequest struct {
	Model             string                  `json:"model"`
	SystemInstruction *GeminiContent          `json:"systemInstruction,omitempty"`
	Contents          []*GeminiContent        `json:"contents"`
	Tools             []*GeminiTool           `json:"tools,omitempty"`
	GenerationConfig  *GeminiGenerationConfig `json:"generationConfig,omitempty"`
}

GeminiRequest represents the structure of a request to the Gemini API. https://ai.google.dev/api/generate-content#v1beta.GenerationConfig

type GeminiRerankRecord ¶

type GeminiRerankRecord struct {
	Id      string  `json:"id,omitempty"`
	Title   string  `json:"title,omitempty"`
	Content string  `json:"content,omitempty"`
	Score   float32 `json:"score,omitempty"`
}

type GeminiThinkingConfig ¶

type GeminiThinkingConfig struct {
	// The number of thoughts tokens that the model should generate.
	ThinkingBudget int `json:"thinkingBudget,omitempty"`

	// Indicates whether to include thoughts in the response. If true, thoughts are returned only when available.
	IncludeThoughts bool `json:"includeThoughts"`
}

https://ai.google.dev/api/generate-content#ThinkingConfig

type GeminiTool ¶

type GeminiTool struct {
	FunctionDeclarations []*GeminiFunctionDeclaration `json:"functionDeclarations"`
}

GeminiTool is a local struct to avoid genai dependency.

type LocalCandidate ¶

type LocalCandidate struct {
	Content      *LocalContent `json:"content"`
	FinishReason string        `json:"finishReason"` // Using string instead of enum
}

LocalCandidate mirrors genai.Candidate

type LocalContent ¶

type LocalContent struct {
	Parts []*LocalPart `json:"parts"`
	Role  string       `json:"role"`
}

LocalContent mirrors genai.Content

type LocalPart ¶

type LocalPart struct {
	Text         *string             `json:"text,omitempty"`
	FunctionCall *GeminiFunctionCall `json:"functionCall,omitempty"`
}

LocalPart is a struct to unmarshal different part types from JSON.

type LocalUsageMetadata ¶

type LocalUsageMetadata struct {
	PromptTokenCount        int32 `json:"promptTokenCount,omitempty"`
	CandidatesTokenCount    int32 `json:"candidatesTokenCount,omitempty"`
	TotalTokenCount         int32 `json:"totalTokenCount,omitempty"`
	CachedContentTokenCount int32 `json:"cachedContentTokenCount,omitempty"`
	ThoughtsTokenCount      int32 `json:"thoughtsTokenCount,omitempty"`
}

LocalUsageMetadata mirrors genai.UsageMetadata

type OpenAIChatMessage ¶

type OpenAIChatMessage struct {
	Role       string                 `json:"role"`
	Content    *string                `json:"content"` // Use pointer to allow for null content
	Name       string                 `json:"name,omitempty"`
	Contents   []OpenAIMessageContent `json:"contents,omitempty"` // Use pointer to allow for null content
	ToolCalls  []ToolCall             `json:"tool_calls,omitempty"`
	ToolCallId string                 `json:"tool_call_id,omitempty"`
	Refusal    string                 `json:"refusal,omitempty"`
}

OpenAIChatMessage mimics the structure of a message in an OpenAI Chat Completion request.

func (*OpenAIChatMessage) GetContent ¶

func (m *OpenAIChatMessage) GetContent() string

type OpenAIChatResponse ¶

type OpenAIChatResponse struct {
	ID          string         `json:"id,omitempty"`
	Created     int64          `json:"created,omitempty"`
	Object      string         `json:"object,omitempty"`
	Model       string         `json:"model,omitempty"`
	Choices     []OpenAIChoice `json:"choices,omitempty"`
	Usage       *Usage         `json:"usage,omitempty"`
	Error       *OpenAIError   `json:"error,omitempty"`
	ServiceTier string         `json:"service_tier,omitempty"`
}

OpenAIChatResponse mimics the structure of an OpenAI Chat Completion response.

func ChatCompleteAPI ¶

func ChatCompleteAPI(ctx context.Context, payload []byte) (OpenAIChatResponse, error)

type OpenAIChoice ¶

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

OpenAIChoice mimics the structure of a choice in an OpenAI Chat Completion response.

type OpenAIEmbeddingData ¶

type OpenAIEmbeddingData struct {
	Object    string    `json:"object"`
	Embedding []float32 `json:"embedding"`
	Index     int       `json:"index"`
}

type OpenAIEmbeddingRequest ¶

type OpenAIEmbeddingRequest struct {
	Input string `json:"input"`
	Model string `json:"model"`
}

type OpenAIEmbeddingResponse ¶

type OpenAIEmbeddingResponse struct {
	Object string                `json:"object"`
	Data   []OpenAIEmbeddingData `json:"data"`
	Model  string                `json:"model"`
	Usage  *Usage                `json:"usage"`
	Error  *OpenAIError          `json:"error,omitempty"`
}

func GetEmbeddingAPI ¶

func GetEmbeddingAPI(ctx context.Context, model, text string) (OpenAIEmbeddingResponse, error)

type OpenAIError ¶

type OpenAIError struct {
	Message string `json:"message"`
	Type    string `json:"type"`
	Param   string `json:"param"`
	Code    string `json:"code,omitempty"`
}

OpenAIError represents the structure of an error in an OpenAI Chat Completion response.

type OpenAIJSONSchema ¶

type OpenAIJSONSchema struct {
	Title                string                       `json:"title,omitempty"`
	Type                 string                       `json:"type,omitempty"` // string, number, object, array, boolean, null
	Description          string                       `protobuf:"bytes,5,opt,name=description,proto3" json:"description,omitempty"`
	Properties           map[string]*OpenAIJSONSchema ``                                                                /* 146-byte string literal not displayed */
	Items                *OpenAIJSONSchema            `protobuf:"bytes,7,opt,name=items,proto3" json:"items,omitempty"` // used for type array
	MinItems             int64                        `protobuf:"varint,8,opt,name=minItems,proto3" json:"minItems,omitempty"`
	UniqueItems          bool                         `protobuf:"varint,9,opt,name=uniqueItems,proto3" json:"uniqueItems,omitempty"`
	ExclusiveMinimum     int64                        `protobuf:"varint,10,opt,name=exclusiveMinimum,proto3" json:"exclusiveMinimum,omitempty"`
	Required             []string                     `protobuf:"bytes,11,rep,name=required,proto3" json:"required,omitempty"`
	AdditionalProperties bool                         `json:"additionalProperties"` // chatgpt required this
	Enum                 []string                     `json:"enum,omitempty"`
}

type OpenAIMessageContent ¶

type OpenAIMessageContent struct {
	Text     string                        `json:"text,omitempty"`
	Type     string                        `json:"type,omitempty"` // text, image_url
	ImageUrl *OpenAIMessageContentImageUrl `json:"image_url,omitempty"`
}

type OpenAIMessageContentImageUrl ¶

type OpenAIMessageContentImageUrl struct {
	Url string `json:"url,omitempty"`
}

type OpenAITool ¶

type OpenAITool struct {
	Type     string    `json:"type"`
	Function *Function `json:"function"`
	Output   string    `json:"output,omitempty"` // our fake
}

OpenAITool mimics the structure of a tool in an OpenAI Chat Completion request.

type PromptTokensDetails ¶

type PromptTokensDetails struct {
	CachedTokens int64 `json:"cached_tokens"`
}

PromptTokensDetails mimics the structure of the prompt_tokens_details field.

type RerankInput ¶

type RerankInput struct {
	Seed    int             `json:"seed,omitempty"`
	TopN    int             `json:"top_n,omitempty"`
	Model   string          `json:"model,omitempty"`
	Query   string          `json:"query,omitempty"`
	Records []*RerankRecord `json:"records,omitempty"`
}

type RerankOutput ¶

type RerankOutput struct {
	Records     []*RerankRecord `json:"records,omitempty"`
	Created     int64           `json:"created,omitempty"`
	DurationMs  int64           `json:"duration_ms"`
	KfpvCostUSD int64           `json:"kfpv_cost_usd"` // 1 usd -> 1000_000_000 kfpvusd
}

func Rerank ¶

func Rerank(ctx context.Context, model, query string, inrecords []*RerankRecord) (RerankOutput, error)

type RerankRecord ¶

type RerankRecord struct {
	Id      string  `json:"id,omitempty"`
	Title   string  `json:"title,omitempty"`
	Content string  `json:"content,omitempty"`
	Score   float32 `json:"score,omitempty"`
}

type RerankingResponse ¶

type RerankingResponse struct {
	Records []*RerankRecord `json:"records,omitempty"`
	Created int64           `json:"created,omitempty"` // sec
	Object  string          `json:"object,omitempty"`
	Model   string          `json:"model,omitempty"`
	Usage   *Usage          `json:"usage,omitempty"`
	Error   *OpenAIError    `json:"error,omitempty"`
}

func RerankAPI ¶

func RerankAPI(ctx context.Context, token string, payload []byte) (RerankingResponse, error)

type ResponseFormat ¶

type ResponseFormat struct {
	Type       string                              `json:"type,omitempty"`
	JSONSchema *header.LLMResponseJSONSchemaFormat `json:"json_schema,omitempty"`
}

ResponseFormat specifies the format of the response.

type ToolCall ¶

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

ToolCall represents a tool call in an OpenAI response.

type ToolFunction ¶

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

ToolFunction represents the function details in a tool call.

type TotalCost ¶

type TotalCost struct {
	USD int64 `json:"usd"` // kfpvusd = usd*1_000_000_000 = fpvusd * 1000
}

type Usage ¶

type Usage struct {
	PromptTokens        int64                `json:"prompt_tokens"`
	CompletionTokens    int64                `json:"completion_tokens"`
	TotalTokens         int64                `json:"total_tokens"`
	PromptTokensDetails *PromptTokensDetails `json:"prompt_tokens_details,omitempty"`

	KFpvCostUSD int64 `json:"kfpv_cost_usd"` // our fields not openai
	CostVND     int64 `json:"cost_vnd"`      // our fields not openai
}

Usage mimics the structure of the usage field in an OpenAI Chat Completion response.

Source Files ¶

View all Source files

?	: This menu
/	: Search site
f or F	: Jump to
y or Y	: Canonical URL

README ¶

AI Router

Features

Installation

Getting Started

Initialization

Usage Examples

1. Chat Completion

2. Text Embedding

3. Rerank

Supported Models

Chat Completion

Text Embedding

Reranking

License

Documentation ¶

Index ¶

Constants ¶

Variables ¶

Functions ¶

func CalculateCost ¶

func CleanString ¶

func FakeBackend ¶

func GetFallbackChatModel ¶

func GetMD5Hash ¶

func Init ¶

func InitAPI ¶

func ToGeminiModel ¶

func ToGeminiRequestJSON ¶

func ToModel ¶

func ToOpenAICompletionJSON ¶

func ToOpenAIModel ¶

Types ¶

type CompletionInput ¶

type CompletionOutput ¶

func Complete ¶

type CompletionReasoning ¶

type Embedding ¶

type EmbeddingOutput ¶

func GetEmbedding ¶

type Function ¶

type GeminiAPIResponse ¶

type GeminiContent ¶

type GeminiEmbeddingContent ¶

type GeminiEmbeddingPart ¶

type GeminiEmbeddingRequest ¶

type GeminiEmbeddingResponse ¶

type GeminiError ¶

type GeminiErrorDetail ¶

type GeminiFunctionCall ¶

type GeminiFunctionDeclaration ¶

type GeminiFunctionResponse ¶

type GeminiGenerationConfig ¶

type GeminiPart ¶

type GeminiPartInlineData ¶

type GeminiRankingResponse ¶

type GeminiRequest ¶

type GeminiRerankRecord ¶

type GeminiThinkingConfig ¶

type GeminiTool ¶

type LocalCandidate ¶

type LocalContent ¶

type LocalPart ¶

type LocalUsageMetadata ¶

type OpenAIChatMessage ¶

func (*OpenAIChatMessage) GetContent ¶

type OpenAIChatResponse ¶

func ChatCompleteAPI ¶

type OpenAIChoice ¶

type OpenAIEmbeddingData ¶

type OpenAIEmbeddingRequest ¶

type OpenAIEmbeddingResponse ¶

func GetEmbeddingAPI ¶

type OpenAIError ¶

type OpenAIJSONSchema ¶

type OpenAIMessageContent ¶

type OpenAIMessageContentImageUrl ¶

type OpenAITool ¶

type PromptTokensDetails ¶

type RerankInput ¶