dashscopego

README

dashscopego

阿里云平台 dashscope api 的 golang 封装 (非官方)

开通DashScope并创建API-KEY

Install:

go get -u github.com/devinyf/dashscopego

功能列表 & Examples:

通义千问

• 大语言模型
• 千问VL(视觉理解模型)
• 千问Audio(音频语言模型)

  Plugins 模型插件

  • pdf解析
  • Python代码解释器
  • 计算器
  • 图片生成

  Function-Call

  • 自定义工具调用

通义万相

• 文本生成图像
• 涂鸦作画
• AnyText图文融合
• 人像风格重绘
• 图像背景生成

Paraformer(语音识别文字)

• 实时语音识别
• 录音文件识别

语音合成

• 文本至语音的实时流式合成

Embedding(通用文本向量)

• 同步接口
• 批处理接口

Langchaingo 集成 (Experimental...)

• 通过 langchiango-agent 调用工具: 通义万象生图
• 通过 langchaingo 调用 function-call 自定义函数

---

基本说明:

import (
	"context"
	"fmt"
	"os"

	"github.com/devinyf/dashscopego"
	"github.com/devinyf/dashscopego/qwen"
)

func main() {
	// 定义 Model,API-KEY, 请求客户端
	model := qwen.QwenTurbo
	token := os.Getenv("DASHSCOPE_API_KEY")

	if token == "" {
		panic("token is empty")
	}

	cli := dashscopego.NewTongyiClient(model, token)

	/*
	 * 开启流式输出:
	 * 通过该 Callback Function 获取流式输出的结果, 如果没有定义该回调函数则默认使用非流式输出
	 * 流式输出结果的 request_id/finish_reason/token_usage 等信息在调用完成后返回的 resp 结果中统一获取
	 */
	streamCallbackFn := func(ctx context.Context, chunk []byte) error {
		// 也可以通过外部定义的 channel 将结果传递出去
		fmt.Print(string(chunk))
		return nil
	}

	// 定义请求内容.
	// 请求具体字段说明请查阅官方文档的 'HTTP调用接口' 部分.
	content := qwen.TextContent{Text: "讲个冷笑话"}

	input := dashscopego.TextInput{
		Messages: []dashscopego.TextMessage{
			{Role: qwen.RoleUser, Content: &content},
		},
	}

	req := &dashscopego.TextRequest{
		Input:       input,             // 请求内容
		StreamingFn: streamCallbackFn,  // 流式输出的回调函数, 默认为 nil, 表示不使用流式输出.
	}

	// 发送请求.
	ctx := context.TODO()
	resp, err := cli.CreateCompletion(ctx, req)
	if err != nil {
		panic(err)
	}

	/*
	获取结果.
	详细字段说明请查阅 'HTTP调用接口 -> 出参描述'.
	如果request中没有定义流式输出的回调函数 StreamingFn, 则使用此方法获取应答内容.
	*/ 
	fmt.Println(resp.Output.Choices[0].Message.Content.ToString())

	// 获取 RequestcID, Token 消耗， 结束标识等信息
	fmt.Println(resp.RequestID)
	fmt.Println(resp.Output.Choices[0].FinishReason)
	fmt.Println(resp.Usage.TotalTokens)
}

Documentation

Index

• Constants
• Variables
• func NewQwenMessage[T qwen.IQwenContent](role string, content T) *qwen.Message[T]
• type AudioInput
• type AudioMessage
• type AudioQwenResponse
• type AudioRequest
• type FileInput
• type FileMessage
• type FileQwenResponse
• type FileRequest
• type TextInput
• type TextMessage
• type TextQwenResponse
• type TextRequest
• type TongyiClient
  • func NewTongyiClient(model string, token string) *TongyiClient
  • func NewTongyiClientIntl(model string, token string) *TongyiClient
  • func (q *TongyiClient) CloseSpeechToTextGeneration() error
  • func (q *TongyiClient) CreateAudioCompletion(ctx context.Context, payload *qwen.Request[*qwen.AudioContentList]) (*AudioQwenResponse, error)
  • func (q *TongyiClient) CreateCompletion(ctx context.Context, payload *qwen.Request[*qwen.TextContent]) (*TextQwenResponse, error)
  • func (q *TongyiClient) CreateEmbedding(ctx context.Context, r *embedding.Request) ([][]float64, int, error)
  • func (q *TongyiClient) CreateFileCompletion(ctx context.Context, payload *qwen.Request[*qwen.FileContentList]) (*FileQwenResponse, error)
  • func (q *TongyiClient) CreateImageGeneration(ctx context.Context, payload *wanx.ImageSynthesisRequest) ([]*wanx.ImgBlob, error)
  • func (q *TongyiClient) CreateSpeechToTextGeneration(ctx context.Context, request *paraformer.Request, reader *bufio.Reader) error
  • func (q *TongyiClient) CreateVLCompletion(ctx context.Context, payload *qwen.Request[*qwen.VLContentList]) (*VLQwenResponse, error)
  • func (q *TongyiClient) CreateVoiceFileToTextGeneration(ctx context.Context, request *paraformer.AsyncTaskRequest) (*paraformer.VoiceFileResponse, error)
  • func (q *TongyiClient) SetUploadCache(uploadCache qwen.UploadCacher) *TongyiClient
• type VLInput
• type VLMessage
• type VLQwenResponse
• type VLRequest
• type WrapMessageError
  • func (e *WrapMessageError) Error() string

Constants

const (
	DashscopeTokenEnvName = "DASHSCOPE_API_KEY" //nolint:gosec
)

Variables

var (
	ErrModelNotSet     = errors.New("model is not set")
	ErrEmptyResponse   = errors.New("empty response")
	ErrImageFilePrefix = errors.New("file prefix is not supported, must be one of: file://, https://, http://")
)

Functions

func NewQwenMessage

func NewQwenMessage[T qwen.IQwenContent](role string, content T) *qwen.Message[T]

Types

type AudioInput

type AudioInput = qwen.Input[*qwen.AudioContentList]

type AudioMessage

type AudioMessage = qwen.Message[*qwen.AudioContentList]

type AudioQwenResponse

type AudioQwenResponse = qwen.OutputResponse[*qwen.AudioContentList]

type AudioRequest

type AudioRequest = qwen.Request[*qwen.AudioContentList]

type FileInput

type FileInput = qwen.Input[*qwen.FileContentList]

type FileMessage

type FileMessage = qwen.Message[*qwen.FileContentList]

type FileQwenResponse

type FileQwenResponse = qwen.OutputResponse[*qwen.TextContent] // PDF 文件解析返回的是纯文本格式.

type FileRequest

type FileRequest = qwen.Request[*qwen.FileContentList]

type TextInput

type TextInput = qwen.Input[*qwen.TextContent]

type TextMessage

type TextMessage = qwen.Message[*qwen.TextContent]

type TextQwenResponse

type TextQwenResponse = qwen.OutputResponse[*qwen.TextContent]

type TextRequest

type TextRequest = qwen.Request[*qwen.TextContent]

type TongyiClient

type TongyiClient struct {
	Model string
	// contains filtered or unexported fields
}

func NewTongyiClient

func NewTongyiClient(model string, token string) *TongyiClient

qwen client with default base url (CN).

func NewTongyiClientIntl

func NewTongyiClientIntl(model string, token string) *TongyiClient

qwen client with international base url.

func (*TongyiClient) CloseSpeechToTextGeneration

func (q *TongyiClient) CloseSpeechToTextGeneration() error

func (*TongyiClient) CreateAudioCompletion

func (q *TongyiClient) CreateAudioCompletion(ctx context.Context, payload *qwen.Request[*qwen.AudioContentList]) (*AudioQwenResponse, error)

func (*TongyiClient) CreateCompletion

func (q *TongyiClient) CreateCompletion(ctx context.Context, payload *qwen.Request[*qwen.TextContent]) (*TextQwenResponse, error)

duplicate: CreateCompletion and CreateVLCompletion are the same but with different payload types. maybe this can be change in the future.

nolint:lll

func (*TongyiClient) CreateEmbedding

func (q *TongyiClient) CreateEmbedding(ctx context.Context, r *embedding.Request) ([][]float64, int, error)

func (*TongyiClient) CreateFileCompletion

func (q *TongyiClient) CreateFileCompletion(ctx context.Context, payload *qwen.Request[*qwen.FileContentList]) (*FileQwenResponse, error)

used for pdf_extracter plugin.

func (*TongyiClient) CreateImageGeneration

func (q *TongyiClient) CreateImageGeneration(ctx context.Context, payload *wanx.ImageSynthesisRequest) ([]*wanx.ImgBlob, error)

TODO: intergrate wanx.Request into qwen.IQwenContent(or should rename to ITongyiContent)

func (*TongyiClient) CreateSpeechToTextGeneration

func (q *TongyiClient) CreateSpeechToTextGeneration(ctx context.Context, request *paraformer.Request, reader *bufio.Reader) error

realtime sppech to text.

func (*TongyiClient) CreateVLCompletion

func (q *TongyiClient) CreateVLCompletion(ctx context.Context, payload *qwen.Request[*qwen.VLContentList]) (*VLQwenResponse, error)

func (*TongyiClient) CreateVoiceFileToTextGeneration

func (q *TongyiClient) CreateVoiceFileToTextGeneration(ctx context.Context, request *paraformer.AsyncTaskRequest) (*paraformer.VoiceFileResponse, error)

voice file to text.

func (*TongyiClient) SetUploadCache

func (q *TongyiClient) SetUploadCache(uploadCache qwen.UploadCacher) *TongyiClient

setup upload cache for uploading files to prevent duplicate upload.

type VLInput

type VLInput = qwen.Input[*qwen.VLContentList]

type VLMessage

type VLMessage = qwen.Message[*qwen.VLContentList]

type VLQwenResponse

type VLQwenResponse = qwen.OutputResponse[*qwen.VLContentList]

type VLRequest

type VLRequest = qwen.Request[*qwen.VLContentList]

type WrapMessageError

type WrapMessageError struct {
	Message string
	Cause   error
}

func (*WrapMessageError) Error

func (e *WrapMessageError) Error() string