Documentation
¶
Overview ¶
Package groq provides a unofficial client for the Groq API.
With specially designed hardware, the Groq API is a super fast way to query open source llms.
API Documentation: https://console.groq.com/docs/quickstart
Features:
- Supports all models from [Groq](https://wow.groq.com/) in a type-safe way. - Supports streaming. - Supports moderation. - Supports audio transcription. - Supports audio translation. - Supports Tool Use. - Supports Function Calling.
Index ¶
- Constants
- func ToSnakeCase(str string) string
- type APIError
- type AudioRequest
- type AudioResponse
- type AudioResponseFormat
- type ChatCompletionChoice
- type ChatCompletionMessage
- type ChatCompletionRequest
- type ChatCompletionResponse
- type ChatCompletionResponseFormat
- type ChatCompletionResponseFormatJSONSchema
- type ChatCompletionResponseFormatType
- type ChatCompletionStream
- type ChatCompletionStreamChoice
- type ChatCompletionStreamChoiceDelta
- type ChatCompletionStreamResponse
- type ChatMessageImageURL
- type ChatMessagePart
- type ChatMessagePartType
- type Client
- func (c *Client) CreateChatCompletion(ctx context.Context, request ChatCompletionRequest) (response ChatCompletionResponse, err error)
- func (c *Client) CreateChatCompletionJSON(ctx context.Context, request ChatCompletionRequest, output any) (err error)
- func (c *Client) CreateChatCompletionStream(ctx context.Context, request ChatCompletionRequest) (stream *ChatCompletionStream, err error)
- func (c *Client) CreateCompletion(ctx context.Context, request CompletionRequest) (response CompletionResponse, err error)
- func (c *Client) CreateCompletionStream(ctx context.Context, request CompletionRequest) (*CompletionStream, error)
- func (c *Client) CreateTranscription(ctx context.Context, request AudioRequest) (response AudioResponse, err error)
- func (c *Client) CreateTranslation(ctx context.Context, request AudioRequest) (response AudioResponse, err error)
- func (c *Client) Moderate(ctx context.Context, request ModerationRequest) (response Moderation, err error)
- type CompletionChoice
- type CompletionRequest
- type CompletionResponse
- type CompletionStream
- type DefaultErrorAccumulator
- type Endpoint
- type ErrChatCompletionInvalidModel
- type ErrChatCompletionStreamNotSupported
- type ErrCompletionRequestPromptTypeNotSupported
- type ErrCompletionStreamNotSupported
- type ErrCompletionUnsupportedModel
- type ErrContentFieldsMisused
- type ErrTooManyEmptyStreamMessages
- type FinishReason
- type Format
- type FunctionCall
- type FunctionDefinition
- type HarmfulCategory
- type ImageURLDetail
- type LogProb
- type LogProbs
- type LogprobResult
- type Model
- type Moderation
- type ModerationRequest
- type Opts
- type PromptAnnotation
- type PromptFilterResult
- type RateLimitHeaders
- type RawResponse
- type ResetTime
- type Role
- type Segments
- type StreamOptions
- type Tool
- type ToolCall
- type ToolChoice
- type ToolFunction
- type ToolType
- type TopLogProbs
- type TranscriptionTimestampGranularity
- type Usage
- type Words
Constants ¶
const ( ChatMessageRoleSystem Role = "system" // ChatMessageRoleSystem is the system chat message role. ChatMessageRoleUser Role = "user" // ChatMessageRoleUser is the user chat message role. ChatMessageRoleAssistant Role = "assistant" // ChatMessageRoleAssistant is the assistant chat message role. ChatMessageRoleFunction Role = "function" // ChatMessageRoleFunction is the function chat message role. ChatMessageRoleTool Role = "tool" // ChatMessageRoleTool is the tool chat message role. ImageURLDetailHigh ImageURLDetail = "high" // ImageURLDetailHigh is the high image url detail. ImageURLDetailLow ImageURLDetail = "low" // ImageURLDetailLow is the low image url detail. ImageURLDetailAuto ImageURLDetail = "auto" // ImageURLDetailAuto is the auto image url detail. ChatMessagePartTypeText ChatMessagePartType = "text" // ChatMessagePartTypeText is the text chat message part type. ChatMessagePartTypeImageURL ChatMessagePartType = "image_url" // ChatMessagePartTypeImageURL is the image url chat message part type. ChatCompletionResponseFormatTypeJSONObject ChatCompletionResponseFormatType = "json_object" // ChatCompletionResponseFormatTypeJSONObject is the json object chat completion response format type. ChatCompletionResponseFormatTypeJSONSchema ChatCompletionResponseFormatType = "json_schema" // ChatCompletionResponseFormatTypeJSONSchema is the json schema chat completion response format type. ChatCompletionResponseFormatTypeText ChatCompletionResponseFormatType = "text" // ChatCompletionResponseFormatTypeText is the text chat completion response format type. ToolTypeFunction ToolType = "function" // ToolTypeFunction is the function tool type. FinishReasonStop FinishReason = "stop" // FinishReasonStop is the stop finish reason. FinishReasonLength FinishReason = "length" // FinishReasonLength is the length finish reason. FinishReasonFunctionCall FinishReason = "function_call" // FinishReasonFunctionCall is the function call finish reason. FinishReasonToolCalls FinishReason = "tool_calls" // FinishReasonToolCalls is the tool calls finish reason. FinishReasonContentFilter FinishReason = "content_filter" // FinishReasonContentFilter is the content filter finish reason. FinishReasonNull FinishReason = "null" // FinishReasonNull is the null finish reason. )
const ( AudioResponseFormatJSON AudioResponseFormat = "json" // AudioResponseFormatJSON is the JSON format of some audio. AudioResponseFormatText AudioResponseFormat = "text" // AudioResponseFormatText is the text format of some audio. AudioResponseFormatSRT AudioResponseFormat = "srt" // AudioResponseFormatSRT is the SRT format of some audio. AudioResponseFormatVerboseJSON AudioResponseFormat = "verbose_json" // AudioResponseFormatVerboseJSON is the verbose JSON format of some audio. AudioResponseFormatVTT AudioResponseFormat = "vtt" // AudioResponseFormatVTT is the VTT format of some audio. TranscriptionTimestampGranularityWord TranscriptionTimestampGranularity = "word" // TranscriptionTimestampGranularityWord is the word timestamp granularity. TranscriptionTimestampGranularitySegment TranscriptionTimestampGranularity = "segment" // TranscriptionTimestampGranularitySegment is the segment timestamp granularity. Llama370B8192 Model = "llama3-70b-8192" // Llama370B8192 is an AI text model provided by Meta. It has 8192 context window. DistilWhisperLargeV3En Model = "distil-whisper-large-v3-en" // DistilWhisperLargeV3En is an AI audio model provided by Hugging Face. It has 448 context window. Gemma7BIt Model = "gemma-7b-it" // Gemma7BIt is an AI text model provided by Google. It has 8192 context window. LlavaV157B4096Preview Model = "llava-v1.5-7b-4096-preview" // LlavaV157B4096Preview is an AI text model provided by Other. It has 4096 context window. Llama3170BVersatile Model = "llama-3.1-70b-versatile" // Llama3170BVersatile is an AI text model provided by Meta. It has 131072 context window. Llama38B8192 Model = "llama3-8b-8192" // Llama38B8192 is an AI text model provided by Meta. It has 8192 context window. Llama318BInstant Model = "llama-3.1-8b-instant" // Llama318BInstant is an AI text model provided by Meta. It has 131072 context window. WhisperLargeV3 Model = "whisper-large-v3" // WhisperLargeV3 is an AI audio model provided by OpenAI. It has 448 context window. Llama3Groq8B8192ToolUsePreview Model = "llama3-groq-8b-8192-tool-use-preview" // Llama3Groq8B8192ToolUsePreview is an AI text model provided by Groq. It has 8192 context window. Gemma29BIt Model = "gemma2-9b-it" // Gemma29BIt is an AI text model provided by Google. It has 8192 context window. Mixtral8X7B32768 Model = "mixtral-8x7b-32768" // Mixtral8X7B32768 is an AI text model provided by Mistral AI. It has 32768 context window. Llama3Groq70B8192ToolUsePreview Model = "llama3-groq-70b-8192-tool-use-preview" // Llama3Groq70B8192ToolUsePreview is an AI text model provided by Groq. It has 8192 context window. LlamaGuard38B Model = "llama-guard-3-8b" // LlamaGuard38B is an AI moderation model provided by Meta. It has 8192 context window. )
const EmptyID id = ""
EmptyID is used to explicitly define an ID with no value.
Variables ¶
This section is empty.
Functions ¶
func ToSnakeCase ¶ added in v0.9.1
ToSnakeCase converts the provided string into snake case using dashes. This is useful for Schema IDs and definitions to be coherent with common JSON Schema examples.
Types ¶
type APIError ¶ added in v0.9.0
type APIError struct {
Code any `json:"code,omitempty"` // Code is the code of the error.
Message string `json:"message"` // Message is the message of the error.
Param *string `json:"param,omitempty"` // Param is the param of the error.
Type string `json:"type"` // Type is the type of the error.
HTTPStatusCode int `json:"-"` // HTTPStatusCode is the status code of the error.
}
APIError provides error information returned by the Groq API.
func (*APIError) UnmarshalJSON ¶ added in v0.9.0
UnmarshalJSON implements the json.Unmarshaler interface.
type AudioRequest ¶ added in v0.9.0
type AudioRequest struct {
Model Model // Model is the model to use for the transcription.
FilePath string // FilePath is either an existing file in your filesystem or a filename representing the contents of Reader.
Reader io.Reader // Reader is an optional io.Reader when you do not want to use an existing file.
Prompt string // Prompt is the prompt for the transcription.
Temperature float32 // Temperature is the temperature for the transcription.
Language string // Language is the language for the transcription. Only for transcription.
Format AudioResponseFormat // Format is the format for the response.
TimestampGranularities []TranscriptionTimestampGranularity // Only for transcription. TimestampGranularities is the timestamp granularities for the transcription.
}
AudioRequest represents a request structure for audio API.
type AudioResponse ¶ added in v0.9.0
type AudioResponse struct {
Task string `json:"task"` // Task is the task of the response.
Language string `json:"language"` // Language is the language of the response.
Duration float64 `json:"duration"` // Duration is the duration of the response.
Segments Segments `json:"segments"` // Segments is the segments of the response.
Words Words `json:"words"` // Words is the words of the response.
Text string `json:"text"` // Text is the text of the response.
Header http.Header // Header is the header of the response.
}
AudioResponse represents a response structure for audio API.
func (*AudioResponse) SetHeader ¶ added in v0.9.0
func (r *AudioResponse) SetHeader(header http.Header)
SetHeader sets the header of the response.
type AudioResponseFormat ¶ added in v0.9.0
type AudioResponseFormat string
AudioResponseFormat is the response format for the audio API.
Response formatted using AudioResponseFormatJSON by default.
string
type ChatCompletionChoice ¶ added in v0.9.0
type ChatCompletionChoice struct {
Index int `json:"index"` // Index is the index of the choice.
// Message is the chat completion message of the choice.
Message ChatCompletionMessage `json:"message"` // Message is the chat completion message of the choice.
// FinishReason is the finish reason of the choice.
//
// stop: API returned complete message,
// or a message terminated by one of the stop sequences provided via the stop parameter
// length: Incomplete model output due to max_tokens parameter or token limit
// function_call: The model decided to call a function
// content_filter: Omitted content due to a flag from our content filters
// null: API response still in progress or incomplete
FinishReason FinishReason `json:"finish_reason"` // FinishReason is the finish reason of the choice.
// LogProbs is the log probs of the choice.
//
// This is basically the probability of the model choosing the token.
LogProbs *LogProbs `json:"logprobs,omitempty"` // LogProbs is the log probs of the choice.
}
ChatCompletionChoice represents the chat completion choice.
type ChatCompletionMessage ¶ added in v0.9.0
type ChatCompletionMessage struct {
Role Role `json:"role"` // Role is the role of the chat completion message.
Content string `json:"content"` // Content is the content of the chat completion message.
MultiContent []ChatMessagePart // MultiContent is the multi content of the chat completion message.
// This property isn't in the official documentation, but it's in
// the documentation for the official library for python:
// - https://github.com/openai/openai-python/blob/main/chatml.md
// - https://github.com/openai/openai-cookbook/blob/main/examples/How_to_count_tokens_with_tiktoken.ipynb
Name string `json:"name,omitempty"`
// FunctionCall setting for Role=assistant prompts this may be set to the function call generated by the model.
FunctionCall *FunctionCall `json:"function_call,omitempty"`
// ToolCalls setting for Role=assistant prompts this may be set to the tool calls generated by the model, such as function calls.
ToolCalls []ToolCall `json:"tool_calls,omitempty"`
// ToolCallID is setting for Role=tool prompts this should be set to the ID given in the assistant's prior request to call a tool.
ToolCallID string `json:"tool_call_id,omitempty"`
}
ChatCompletionMessage represents the chat completion message.
func (ChatCompletionMessage) MarshalJSON ¶ added in v0.9.0
func (m ChatCompletionMessage) MarshalJSON() ([]byte, error)
MarshalJSON method implements the json.Marshaler interface.
func (*ChatCompletionMessage) UnmarshalJSON ¶ added in v0.9.0
func (m *ChatCompletionMessage) UnmarshalJSON(bs []byte) (err error)
UnmarshalJSON method implements the json.Unmarshaler interface.
type ChatCompletionRequest ¶ added in v0.9.0
type ChatCompletionRequest struct {
Model Model `json:"model"` // Model is the model of the chat completion request.
Messages []ChatCompletionMessage `json:"messages"` // Messages is the messages of the chat completion request. These act as the prompt for the model.
MaxTokens int `json:"max_tokens,omitempty"` // MaxTokens is the max tokens of the chat completion request.
Temperature float32 `json:"temperature,omitempty"` // Temperature is the temperature of the chat completion request.
TopP float32 `json:"top_p,omitempty"` // TopP is the top p of the chat completion request.
N int `json:"n,omitempty"` // N is the n of the chat completion request.
Stream bool `json:"stream,omitempty"` // Stream is the stream of the chat completion request.
Stop []string `json:"stop,omitempty"` // Stop is the stop of the chat completion request.
PresencePenalty float32 `json:"presence_penalty,omitempty"` // PresencePenalty is the presence penalty of the chat completion request.
ResponseFormat *ChatCompletionResponseFormat `json:"response_format,omitempty"` // ResponseFormat is the response format of the chat completion request.
Seed *int `json:"seed,omitempty"` // Seed is the seed of the chat completion request.
FrequencyPenalty float32 `json:"frequency_penalty,omitempty"` // FrequencyPenalty is the frequency penalty of the chat completion request.
LogitBias map[string]int `json:"logit_bias,omitempty"` // LogitBias is must be a token id string (specified by their token ID in the tokenizer), not a word string. incorrect: `"logit_bias":{ "You": 6}`, correct: `"logit_bias":{"1639": 6}` refs: https://platform.openai.com/docs/api-reference/chat/create#chat/create-logit_bias
LogProbs bool `json:"logprobs,omitempty"` // LogProbs indicates whether to return log probabilities of the output tokens or not. If true, returns the log probabilities of each output token returned in the content of message. This option is currently not available on the gpt-4-vision-preview model.
TopLogProbs int `json:"top_logprobs,omitempty"` // TopLogProbs is an integer between 0 and 5 specifying the number of most likely tokens to return at each token position, each with an associated log probability. logprobs must be set to true if this parameter is used.
User string `json:"user,omitempty"` // User is the user of the chat completion request.
Tools []Tool `json:"tools,omitempty"` // Tools is the tools of the chat completion request.
ToolChoice any `json:"tool_choice,omitempty"` // This can be either a string or an ToolChoice object.
StreamOptions *StreamOptions `json:"stream_options,omitempty"` // Options for streaming response. Only set this when you set stream: true.
ParallelToolCalls any `json:"parallel_tool_calls,omitempty"` // Disable the default behavior of parallel tool calls by setting it: false.
}
ChatCompletionRequest represents a request structure for the chat completion API.
type ChatCompletionResponse ¶ added in v0.9.0
type ChatCompletionResponse struct {
ID string `json:"id"` // ID is the id of the response.
Object string `json:"object"` // Object is the object of the response.
Created int64 `json:"created"` // Created is the created time of the response.
Model string `json:"model"` // Model is the model of the response.
Choices []ChatCompletionChoice `json:"choices"` // Choices is the choices of the response.
Usage Usage `json:"usage"` // Usage is the usage of the response.
SystemFingerprint string `json:"system_fingerprint"` // SystemFingerprint is the system fingerprint of the response.
http.Header // Header is the header of the response.
}
ChatCompletionResponse represents a response structure for chat completion API.
func (*ChatCompletionResponse) SetHeader ¶ added in v0.9.0
func (r *ChatCompletionResponse) SetHeader(h http.Header)
SetHeader sets the header of the response.
type ChatCompletionResponseFormat ¶ added in v0.9.0
type ChatCompletionResponseFormat struct {
// Type is the type of the chat completion response format.
Type ChatCompletionResponseFormatType `json:"type,omitempty"`
// JSONSchema is the json schema of the chat completion response format.
JSONSchema *ChatCompletionResponseFormatJSONSchema `json:"json_schema,omitempty"`
}
ChatCompletionResponseFormat is the chat completion response format.
type ChatCompletionResponseFormatJSONSchema ¶ added in v0.9.0
type ChatCompletionResponseFormatJSONSchema struct {
// Name is the name of the chat completion response format json schema.
//
// it is used to further identify the schema in the response.
Name string `json:"name"`
// response format json schema.
// Description is the description of the chat completion response format
// json schema.
Description string `json:"description,omitempty"`
// description of the chat completion response format json schema.
// Schema is the schema of the chat completion response format json schema.
Schema schema `json:"schema"`
// Strict determines whether to enforce the schema upon the generated
// content.
Strict bool `json:"strict"`
}
ChatCompletionResponseFormatJSONSchema is the chat completion response format json schema.
type ChatCompletionResponseFormatType ¶ added in v0.9.0
type ChatCompletionResponseFormatType string
ChatCompletionResponseFormatType is the chat completion response format type.
string
type ChatCompletionStream ¶ added in v0.9.0
type ChatCompletionStream struct {
// contains filtered or unexported fields
}
ChatCompletionStream is a stream of ChatCompletionStreamResponse.
Note: Perhaps it is more elegant to abstract Stream using generics.
type ChatCompletionStreamChoice ¶ added in v0.9.0
type ChatCompletionStreamChoice struct {
Index int `json:"index"`
Delta ChatCompletionStreamChoiceDelta `json:"delta"`
FinishReason FinishReason `json:"finish_reason"`
}
ChatCompletionStreamChoice represents a response structure for chat completion API.
type ChatCompletionStreamChoiceDelta ¶ added in v0.9.0
type ChatCompletionStreamChoiceDelta struct {
Content string `json:"content,omitempty"`
Role string `json:"role,omitempty"`
FunctionCall *FunctionCall `json:"function_call,omitempty"`
ToolCalls []ToolCall `json:"tool_calls,omitempty"`
}
ChatCompletionStreamChoiceDelta represents a response structure for chat completion API.
type ChatCompletionStreamResponse ¶ added in v0.9.0
type ChatCompletionStreamResponse struct {
ID string `json:"id"` // ID is the identifier for the chat completion stream response.
Object string `json:"object"` // Object is the object type of the chat completion stream response.
Created int64 `json:"created"` // Created is the creation time of the chat completion stream response.
Model Model `json:"model"` // Model is the model used for the chat completion stream response.
Choices []ChatCompletionStreamChoice `json:"choices"` // Choices is the choices for the chat completion stream response.
SystemFingerprint string `json:"system_fingerprint"` // SystemFingerprint is the system fingerprint for the chat completion stream response.
PromptAnnotations []PromptAnnotation `json:"prompt_annotations,omitempty"` // PromptAnnotations is the prompt annotations for the chat completion stream response.
PromptFilterResults []PromptFilterResult `json:"prompt_filter_results,omitempty"` // PromptFilterResults is the prompt filter results for the chat completion stream response.
// Usage is an optional field that will only be present when you set stream_options: {"include_usage": true} in your request.
//
// When present, it contains a null value except for the last chunk which contains the token usage statistics
// for the entire request.
Usage *Usage `json:"usage,omitempty"`
}
ChatCompletionStreamResponse represents a response structure for chat completion API.
type ChatMessageImageURL ¶ added in v0.9.0
type ChatMessageImageURL struct {
URL string `json:"url,omitempty"` // URL is the url of the image.
Detail ImageURLDetail `json:"detail,omitempty"` // Detail is the detail of the image url.
}
ChatMessageImageURL represents the chat message image url.
type ChatMessagePart ¶ added in v0.9.0
type ChatMessagePart struct {
Type ChatMessagePartType `json:"type,omitempty"`
Text string `json:"text,omitempty"`
ImageURL *ChatMessageImageURL `json:"image_url,omitempty"`
}
ChatMessagePart represents the chat message part of a chat completion message.
type ChatMessagePartType ¶ added in v0.9.0
type ChatMessagePartType string
ChatMessagePartType is the chat message part type.
string
type Client ¶
type Client struct {
EmptyMessagesLimit uint // EmptyMessagesLimit is the limit for the empty messages.
// contains filtered or unexported fields
}
Client is a Groq api client.
func (*Client) CreateChatCompletion ¶ added in v0.9.0
func (c *Client) CreateChatCompletion( ctx context.Context, request ChatCompletionRequest, ) (response ChatCompletionResponse, err error)
CreateChatCompletion method is an API call to create a chat completion.
func (*Client) CreateChatCompletionJSON ¶ added in v0.9.1
func (c *Client) CreateChatCompletionJSON( ctx context.Context, request ChatCompletionRequest, output any, ) (err error)
CreateChatCompletionJSON method is an API call to create a chat completion w/ object output.
func (*Client) CreateChatCompletionStream ¶ added in v0.9.0
func (c *Client) CreateChatCompletionStream( ctx context.Context, request ChatCompletionRequest, ) (stream *ChatCompletionStream, err error)
CreateChatCompletionStream method is an API call to create a chat completion w/ streaming support.
If set, tokens will be sent as data-only server-sent events as they become available, with the stream terminated by a data: [DONE] message.
func (*Client) CreateCompletion ¶ added in v0.9.0
func (c *Client) CreateCompletion( ctx context.Context, request CompletionRequest, ) (response CompletionResponse, err error)
CreateCompletion call the Groq API to create a completion. This is the main endpoint of the API. Returns new text as well as, if requested, the probabilities over each alternative token at each position.
If using a fine-tuned model, simply provide the model's ID in the CompletionRequest object, and the server will use the model's parameters to generate the completion.
func (*Client) CreateCompletionStream ¶ added in v0.9.0
func (c *Client) CreateCompletionStream( ctx context.Context, request CompletionRequest, ) (*CompletionStream, error)
CreateCompletionStream — API call to create a completion w/ streaming support.
Recv receives a response from the stream. It sets whether to stream back partial progress.
If set, tokens will be sent as data-only server-sent events as they become available, with the stream terminated by a data: [DONE] message.
func (*Client) CreateTranscription ¶ added in v0.9.0
func (c *Client) CreateTranscription( ctx context.Context, request AudioRequest, ) (response AudioResponse, err error)
CreateTranscription calls the transcriptions endpoint with the given request.
Returns transcribed text in the response_format specified in the request.
func (*Client) CreateTranslation ¶ added in v0.9.0
func (c *Client) CreateTranslation( ctx context.Context, request AudioRequest, ) (response AudioResponse, err error)
CreateTranslation calls the translations endpoint with the given request.
Returns the translated text in the response_format specified in the request.
func (*Client) Moderate ¶ added in v0.9.0
func (c *Client) Moderate( ctx context.Context, request ModerationRequest, ) (response Moderation, err error)
Moderate — perform a moderation api call over a string. Input can be an array or slice but a string will reduce the complexity.
type CompletionChoice ¶ added in v0.9.0
type CompletionChoice struct {
Text string `json:"text"` // Text is the text of the completion.
Index int `json:"index"` // Index is the index of the completion.
FinishReason string `json:"finish_reason"` // FinishReason is the finish reason of the completion.
LogProbs LogprobResult `json:"logprobs"` // LogProbs is the log probabilities of the completion.
}
CompletionChoice represents one of possible completions.
type CompletionRequest ¶ added in v0.9.0
type CompletionRequest struct {
Model Model `json:"model"` // Model is the model to use for the completion.
Prompt any `json:"prompt,omitempty"` // Prompt is the prompt for the completion.
BestOf int `json:"best_of,omitempty"` // BestOf is the number of completions to generate.
Echo bool `json:"echo,omitempty"` // Echo is whether to echo back the prompt in the completion.
FrequencyPenalty float32 `json:"frequency_penalty,omitempty"` // FrequencyPenalty is the frequency penalty for the completion.
LogitBias map[string]int `json:"logit_bias,omitempty"` // LogitBias is must be a token id string (specified by their token ID in the tokenizer), not a word string. incorrect: `"logit_bias":{"You": 6}`, correct: `"logit_bias":{"1639": 6}` refs: https://platform.openai.com/docs/api-reference/completions/create#completions/create-logit_bias
LogProbs int `json:"logprobs,omitempty"` // LogProbs is whether to include the log probabilities in the response.
MaxTokens int `json:"max_tokens,omitempty"` // MaxTokens is the maximum number of tokens to generate.
N int `json:"n,omitempty"` // N is the number of completions to generate.
PresencePenalty float32 `json:"presence_penalty,omitempty"` // PresencePenalty is the presence penalty for the completion.
Seed *int `json:"seed,omitempty"` // Seed is the seed for the completion.
Stop []string `json:"stop,omitempty"` // Stop is the stop sequence for the completion.
Stream bool `json:"stream,omitempty"` // Stream is whether to stream the response.
Suffix string `json:"suffix,omitempty"` // Suffix is the suffix for the completion.
Temperature float32 `json:"temperature,omitempty"` // Temperature is the temperature for the completion.
TopP float32 `json:"top_p,omitempty"` // TopP is the top p for the completion.
User string `json:"user,omitempty"` // User is the user for the completion.
}
CompletionRequest represents a request structure for completion API.
type CompletionResponse ¶ added in v0.9.0
type CompletionResponse struct {
ID string `json:"id"` // ID is the ID of the completion.
Object string `json:"object"` // Object is the object of the completion.
Created int64 `json:"created"` // Created is the created time of the completion.
Model Model `json:"model"` // Model is the model of the completion.
Choices []CompletionChoice `json:"choices"` // Choices is the choices of the completion.
Usage Usage `json:"usage"` // Usage is the usage of the completion.
Header http.Header // Header is the header of the response.
}
CompletionResponse represents a response structure for completion API.
func (*CompletionResponse) SetHeader ¶ added in v0.9.0
func (r *CompletionResponse) SetHeader(header http.Header)
SetHeader sets the header of the response.
type CompletionStream ¶ added in v0.9.0
type CompletionStream struct {
// contains filtered or unexported fields
}
CompletionStream is a stream of completions.
type DefaultErrorAccumulator ¶ added in v0.9.0
type DefaultErrorAccumulator struct {
Buffer errorBuffer
}
DefaultErrorAccumulator is a default implementation of ErrorAccumulator
func (*DefaultErrorAccumulator) Bytes ¶ added in v0.9.0
func (e *DefaultErrorAccumulator) Bytes() (errBytes []byte)
Bytes returns the bytes of the error accumulator.
func (*DefaultErrorAccumulator) Write ¶ added in v0.9.0
func (e *DefaultErrorAccumulator) Write(p []byte) error
Write writes bytes to the error accumulator.
type Endpoint ¶ added in v0.9.0
type Endpoint string
Endpoint is the endpoint for the groq api. string
type ErrChatCompletionInvalidModel ¶ added in v0.9.0
ErrChatCompletionInvalidModel is an error that occurs when the model is not supported with the CreateChatCompletion method.
func (ErrChatCompletionInvalidModel) Error ¶ added in v0.9.0
func (e ErrChatCompletionInvalidModel) Error() string
Error implements the error interface.
type ErrChatCompletionStreamNotSupported ¶ added in v0.9.0
type ErrChatCompletionStreamNotSupported struct {
// contains filtered or unexported fields
}
ErrChatCompletionStreamNotSupported is an error that occurs when streaming is not supported with the CreateChatCompletionStream method.
func (ErrChatCompletionStreamNotSupported) Error ¶ added in v0.9.0
func (e ErrChatCompletionStreamNotSupported) Error() string
Error implements the error interface.
type ErrCompletionRequestPromptTypeNotSupported ¶ added in v0.9.0
type ErrCompletionRequestPromptTypeNotSupported struct{}
ErrCompletionRequestPromptTypeNotSupported is an error that occurs when the type of CompletionRequest.Prompt only supports string and []string.
func (ErrCompletionRequestPromptTypeNotSupported) Error ¶ added in v0.9.0
func (e ErrCompletionRequestPromptTypeNotSupported) Error() string
Error implements the error interface.
type ErrCompletionStreamNotSupported ¶ added in v0.9.0
type ErrCompletionStreamNotSupported struct{}
ErrCompletionStreamNotSupported is an error that occurs when streaming is not supported with the CreateCompletionStream method.
func (ErrCompletionStreamNotSupported) Error ¶ added in v0.9.0
func (e ErrCompletionStreamNotSupported) Error() string
Error implements the error interface.
type ErrCompletionUnsupportedModel ¶ added in v0.9.0
type ErrCompletionUnsupportedModel struct{ Model Model }
ErrCompletionUnsupportedModel is an error that occurs when the model is not supported with the CreateCompletion method.
func (ErrCompletionUnsupportedModel) Error ¶ added in v0.9.0
func (e ErrCompletionUnsupportedModel) Error() string
Error implements the error interface.
type ErrContentFieldsMisused ¶ added in v0.9.0
type ErrContentFieldsMisused struct {
// contains filtered or unexported fields
}
ErrContentFieldsMisused is an error that occurs when both Content and MultiContent properties are set.
func (ErrContentFieldsMisused) Error ¶ added in v0.9.0
func (e ErrContentFieldsMisused) Error() string
Error implements the error interface.
type ErrTooManyEmptyStreamMessages ¶ added in v0.9.0
type ErrTooManyEmptyStreamMessages struct{}
ErrTooManyEmptyStreamMessages is returned when the stream has sent too many empty messages.
func (ErrTooManyEmptyStreamMessages) Error ¶ added in v0.9.0
func (e ErrTooManyEmptyStreamMessages) Error() string
Error returns the error message.
type FinishReason ¶ added in v0.9.0
type FinishReason string
FinishReason is the finish reason. string
func (FinishReason) MarshalJSON ¶ added in v0.9.0
func (r FinishReason) MarshalJSON() ([]byte, error)
MarshalJSON implements the json.Marshaler interface.
type FunctionCall ¶ added in v0.9.0
type FunctionCall struct {
Name string `json:"name,omitempty"` // Name is the name of the function call.
Arguments string `json:"arguments,omitempty"` // Arguments is the arguments of the function call in JSON format.
}
FunctionCall represents a function call.
type FunctionDefinition ¶ added in v0.9.0
type FunctionDefinition struct {
Name string `json:"name"` // Name is the name of the function.
Description string `json:"description,omitempty"` // Description is the description of the function.
Strict bool `json:"strict,omitempty"` // Strict is the strict of the function.
// Parameters is an object describing the function.
// You can pass json.RawMessage to describe the schema,
// or you can pass in a struct which serializes to the proper JSON schema.
// The jsonschema package is provided for convenience, but you should
// consider another specialized library if you require more complex schemas.
Parameters any `json:"parameters"`
}
FunctionDefinition represents the function definition.
type HarmfulCategory ¶ added in v0.9.0
type HarmfulCategory string
HarmfulCategory is a category of harmful content.
[Llama Guard 3](https://llama.meta.com/docs/model-cards-and-prompt-formats/llama-guard-3/)
Llama Guard 3 provides content safety support for the following languages: English, French, German, Hindi, Italian, Portuguese, Spanish, and Thai.
string
const ( // ModerationTextStable is the text-moderation-stable model. ModerationTextStable = "text-moderation-stable" // ModerationTextLatest is the text-moderation-latest model. ModerationTextLatest = "text-moderation-latest" // CategoryViolentCrimes (S1) is the violent crimes category. Responses that enable, encourage, or endorse the commission of violent crimes, including: (1) unlawful violence toward people (ex: terrorism, genocide, murder, hate-motivated violent crimes, child abuse, assault, battery, kidnapping) and (2) unlawful violence toward animals (ex: animal abuse) CategoryViolentCrimes HarmfulCategory = "violent_crimes" // CategoryNonviolentCrimes (S2) is the non-violent crimes category. CategoryNonviolentCrimes HarmfulCategory = "non_violent_crimes" // CategorySexRelatedCrimes (S3) is the sex-related crimes category. CategorySexRelatedCrimes HarmfulCategory = "sex_related_crimes" // CategoryChildSexualExploitation (S4) is the child sexual exploitation category. Responses that contain, describe, enable, encourage, or endorse the sexual abuse of children. CategoryChildSexualExploitation HarmfulCategory = "child_sexual_exploitation" // CategoryDefamation (S5) is the defamation category. Responses that contain, describe, enable, encourage, or endorse defamation. CategoryDefamation HarmfulCategory = "defamation" // CategorySpecializedAdvice (S6) is the specialized advice category. Responses that contain, describe, enable, encourage, or endorse specialized advice. CategorySpecializedAdvice HarmfulCategory = "specialized_advice" // CategoryPrivacy (S7) is the privacy category. Responses that contain, describe, enable, encourage, or endorse privacy. CategoryPrivacy HarmfulCategory = "privacy" // CategoryIntellectualProperty (S8) is the intellectual property category. Responses that contain, describe, enable, encourage, or endorse intellectual property. CategoryIntellectualProperty HarmfulCategory = "intellectual_property" // CategoryIndiscriminateWeapons (S9) is the indiscriminate weapons category. Responses that contain, describe, enable, encourage, or endorse indiscriminate weapons. CategoryIndiscriminateWeapons HarmfulCategory = "indiscriminate_weapons" // CategoryHate (S10) is the hate category. Responses that contain, describe, enable, encourage, or endorse hate. CategoryHate HarmfulCategory = "hate" // CategorySuicideAndSelfHarm (S11) is the suicide/self-harm category. Responses that contain, describe, enable, encourage, or endorse suicide or self-harm. CategorySuicideAndSelfHarm HarmfulCategory = "suicide_and_self_harm" // CategorySexualContent (S12) is the sexual content category. Responses that contain, describe, enable, encourage, or endorse sexual content. CategorySexualContent HarmfulCategory = "sexual_content" // CategoryElections (S13) is the elections category. Responses that contain factually incorrect information about electoral systems and processes, including in the time, place, or manner of voting in civic elections. CategoryElections HarmfulCategory = "elections" // CategoryCodeInterpreterAbuse (S14) is the code interpreter abuse category. Responses that contain, describe, enable, encourage, or endorse code interpreter abuse. CategoryCodeInterpreterAbuse HarmfulCategory = "code_interpreter_abuse" )
type ImageURLDetail ¶ added in v0.9.0
type ImageURLDetail string
ImageURLDetail is the image url detail.
string
type LogProb ¶ added in v0.9.0
type LogProb struct {
Token string `json:"token"` // Token is the token of the log prob.
LogProb float64 `json:"logprob"` // LogProb is the log prob of the log prob.
Bytes []byte `json:"bytes,omitempty"` // Omitting the field if it is null
TopLogProbs []TopLogProbs `json:"top_logprobs"` // TopLogProbs is a list of the most likely tokens and their log probability, at this token position. In rare cases, there may be fewer than the number of requested top_logprobs returned.
}
LogProb represents the probability information for a token.
type LogProbs ¶ added in v0.9.0
type LogProbs struct {
Content []LogProb `json:"content"` // Content is a list of message content tokens with log probability information.
}
LogProbs is the top-level structure containing the log probability information.
type LogprobResult ¶ added in v0.9.0
type LogprobResult struct {
Tokens []string `json:"tokens"` // Tokens is the tokens of the completion.
TokenLogprobs []float32 `json:"token_logprobs"` // TokenLogprobs is the token log probabilities of the completion.
TopLogprobs []map[string]float32 `json:"top_logprobs"` // TopLogprobs is the top log probabilities of the completion.
TextOffset []int `json:"text_offset"` // TextOffset is the text offset of the completion.
}
LogprobResult represents logprob result of Choice.
type Model ¶ added in v0.9.0
type Model string
Model is the type for models present on the groq api. string
type Moderation ¶ added in v0.9.0
type Moderation struct {
Categories []HarmfulCategory `json:"categories"` // Categories is the categories of the result.
Flagged bool `json:"flagged"` // Flagged is the flagged of the result.
}
Moderation represents one of possible moderation results.
type ModerationRequest ¶ added in v0.9.0
type ModerationRequest struct {
Input string `json:"input,omitempty"` // Input is the input text to be moderated.
Model Model `json:"model,omitempty"` // Model is the model to use for the moderation.
}
ModerationRequest represents a request structure for moderation API.
type Opts ¶ added in v0.9.0
type Opts func(*Client)
Opts is a function that sets options for a Groq client.
func WithBaseURL ¶ added in v0.9.0
WithBaseURL sets the base URL for the Groq client.
func WithClient ¶
WithClient sets the client for the Groq client.
func WithLogger ¶ added in v0.9.0
WithLogger sets the logger for the Groq client.
type PromptAnnotation ¶ added in v0.9.0
type PromptAnnotation struct {
PromptIndex int `json:"prompt_index,omitempty"`
}
PromptAnnotation represents the prompt annotation.
type PromptFilterResult ¶ added in v0.9.0
type PromptFilterResult struct {
Index int `json:"index"`
}
PromptFilterResult represents a response structure for chat completion API.
type RateLimitHeaders ¶ added in v0.9.0
type RateLimitHeaders struct {
LimitRequests int `json:"x-ratelimit-limit-requests"` // LimitRequests is the limit requests of the rate limit headers.
LimitTokens int `json:"x-ratelimit-limit-tokens"` // LimitTokens is the limit tokens of the rate limit headers.
RemainingRequests int `json:"x-ratelimit-remaining-requests"` // RemainingRequests is the remaining requests of the rate limit headers.
RemainingTokens int `json:"x-ratelimit-remaining-tokens"` // RemainingTokens is the remaining tokens of the rate limit headers.
ResetRequests ResetTime `json:"x-ratelimit-reset-requests"` // ResetRequests is the reset requests of the rate limit headers.
ResetTokens ResetTime `json:"x-ratelimit-reset-tokens"` // ResetTokens is the reset tokens of the rate limit headers.
}
RateLimitHeaders struct represents Groq rate limits headers.
type RawResponse ¶ added in v0.9.0
type RawResponse struct {
io.ReadCloser
http.Header
}
RawResponse is a response from the raw endpoint.
type ResetTime ¶ added in v0.9.0
type ResetTime string
ResetTime is a time.Time wrapper for the rate limit reset time. string
type Role ¶ added in v0.9.0
type Role string
Role is the role of the chat completion message.
string
type Segments ¶ added in v0.9.0
type Segments []struct {
ID int `json:"id"` // ID is the ID of the segment.
Seek int `json:"seek"` // Seek is the seek of the segment.
Start float64 `json:"start"` // Start is the start of the segment.
End float64 `json:"end"` // End is the end of the segment.
Text string `json:"text"` // Text is the text of the segment.
Tokens []int `json:"tokens"` // Tokens is the tokens of the segment.
Temperature float64 `json:"temperature"` // Temperature is the temperature of the segment.
AvgLogprob float64 `json:"avg_logprob"` // AvgLogprob is the avg log prob of the segment.
CompressionRatio float64 `json:"compression_ratio"` // CompressionRatio is the compression ratio of the segment.
NoSpeechProb float64 `json:"no_speech_prob"` // NoSpeechProb is the no speech prob of the segment.
Transient bool `json:"transient"` // Transient is the transient of the segment.
}
Segments is the segments of the response.
type StreamOptions ¶ added in v0.9.0
type StreamOptions struct {
// If set, an additional chunk will be streamed before the data: [DONE] message.
// The usage field on this chunk shows the token usage statistics for the entire request,
// and the choices field will always be an empty array.
// All other chunks will also include a usage field, but with a null value.
IncludeUsage bool `json:"include_usage,omitempty"`
}
StreamOptions represents the stream options.
type Tool ¶ added in v0.9.0
type Tool struct {
Type ToolType `json:"type"` // Type is the type of the tool.
Function *FunctionDefinition `json:"function,omitempty"` // Function is the function of the tool.
}
Tool represents the tool.
type ToolCall ¶ added in v0.9.0
type ToolCall struct {
// Index is not nil only in chat completion chunk object
Index *int `json:"index,omitempty"` // Index is the index of the tool call.
ID string `json:"id"` // ID is the id of the tool call.
Type ToolType `json:"type"` // Type is the type of the tool call.
Function FunctionCall `json:"function"` // Function is the function of the tool call.
}
ToolCall represents a tool call.
type ToolChoice ¶ added in v0.9.0
type ToolChoice struct {
Type ToolType `json:"type"` // Type is the type of the tool choice.
Function ToolFunction `json:"function,omitempty"` // Function is the function of the tool choice.
}
ToolChoice represents the tool choice.
type ToolFunction ¶ added in v0.9.0
type ToolFunction struct {
Name string `json:"name"` // Name is the name of the tool function.
}
ToolFunction represents the tool function.
type TopLogProbs ¶ added in v0.9.0
type TopLogProbs struct {
Token string `json:"token"` // Token is the token of the top log probs.
LogProb float64 `json:"logprob"` // LogProb is the log prob of the top log probs.
Bytes []byte `json:"bytes,omitempty"` // Bytes is the bytes of the top log probs.
}
TopLogProbs represents the top log probs.
type TranscriptionTimestampGranularity ¶ added in v0.9.0
type TranscriptionTimestampGranularity string
TranscriptionTimestampGranularity is the timestamp granularity for the transcription.
string
Source Files
¶
Directories
¶
| Path | Synopsis |
|---|---|
|
examples
|
|
|
json-chat
command
Package main demonstrates an example application of groq-go.
|
Package main demonstrates an example application of groq-go. |
|
llava-blind
command
Package main demonstrates an example application of groq-go.
|
Package main demonstrates an example application of groq-go. |
|
terminal-chat
command
Package main demonstrates how to use groq-go to create a chat application using the groq api accessable through the terminal.
|
Package main demonstrates how to use groq-go to create a chat application using the groq api accessable through the terminal. |
|
extensions
|
|
|
toolhouse
module
|
|
|
internal
|
|
|
test
Package test contains test helpers.
|
Package test contains test helpers. |
|
scripts
|
|
|
models
module
|