zlei6/ragflow
1
0
Fork 0
mirror of https://github.com/infiniflow/ragflow.git synced 2026-09-08 10:14:35 +08:00
Code Issues Packages Projects Releases Wiki Activity

Go: add stats (#17061)

Browse Source 
### Summary

Add LLM token stats framework

---------

Signed-off-by: Jin Hai <haijin.chn@gmail.com>
This commit is contained in:
Jin Hai
2026-07-18 21:02:07 +08:00
committed by GitHub
parent cafbbd467d
commit b8d06d02e6
144 changed files with 1578 additions and 1280 deletions
Download Patch File Download Diff File Expand all files Collapse all files
internal/entity/models/xinference.go
+11 -10
View File
@@ -25,6 +25,7 @@ import (
"net/http"
"os"
"path/filepath"
"ragflow/internal/common"
"strconv"
"strings"
"sync"
@@ -142,7 +143,7 @@ func buildXinferenceChatBody(modelName string, messages []Message, stream bool,
}
// ChatWithMessages sends multiple messages with roles and returns the response.
func (x *XinferenceModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig) (*ChatResponse, error) {
func (x *XinferenceModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage) (*ChatResponse, error) {
if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
return nil, err
}
@@ -212,7 +213,7 @@ func (x *XinferenceModel) ChatWithMessages(modelName string, messages []Message,
}
// ChatStreamlyWithSender sends messages and streams response via sender.
func (x *XinferenceModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, sender func(*string, *string) error) error {
func (x *XinferenceModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
return err
}
@@ -341,7 +342,7 @@ type xinferenceEmbeddingResponse struct {
}
// Embed POSTs the input texts to the tenant's Xinference
func (x *XinferenceModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig) ([]EmbeddingData, error) {
func (x *XinferenceModel) Embed(modelName *string, texts []string, apiConfig *APIConfig, embeddingConfig *EmbeddingConfig, modelUsage *common.ModelUsage) ([]EmbeddingData, error) {
if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
return nil, err
}
@@ -449,7 +450,7 @@ type xinferenceRerankResponse struct {
// in the API's ranking order. Caller may sort by Index to recover
// original input order. Xinference rerank models are launched with
// --model-type rerank and exposed under the OpenAI-compatible base URL.
func (x *XinferenceModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig) (*RerankResponse, error) {
func (x *XinferenceModel) Rerank(modelName *string, query string, documents []string, apiConfig *APIConfig, rerankConfig *RerankConfig, modelUsage *common.ModelUsage) (*RerankResponse, error) {
if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
return nil, err
}
@@ -540,7 +541,7 @@ func (x *XinferenceModel) Rerank(modelName *string, query string, documents []st
return &rerankResponse, nil
}
func (x *XinferenceModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig) (*ASRResponse, error) {
func (x *XinferenceModel) TranscribeAudio(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage) (*ASRResponse, error) {
if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
return nil, err
}
@@ -649,11 +650,11 @@ func (x *XinferenceModel) TranscribeAudio(modelName *string, file *string, apiCo
}, nil
}
func (x *XinferenceModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, sender func(*string, *string) error) error {
func (x *XinferenceModel) TranscribeAudioWithSender(modelName *string, file *string, apiConfig *APIConfig, asrConfig *ASRConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
return fmt.Errorf("%s, no such method", x.Name())
}
func (x *XinferenceModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig) (*TTSResponse, error) {
func (x *XinferenceModel) AudioSpeech(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage) (*TTSResponse, error) {
if err := x.baseModel.APIConfigCheck(apiConfig); err != nil {
return nil, err
}
@@ -715,15 +716,15 @@ func (x *XinferenceModel) AudioSpeech(modelName *string, audioContent *string, a
return &TTSResponse{Audio: body}, nil
}
func (x *XinferenceModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, sender func(*string, *string) error) error {
func (x *XinferenceModel) AudioSpeechWithSender(modelName *string, audioContent *string, apiConfig *APIConfig, ttsConfig *TTSConfig, modelUsage *common.ModelUsage, sender func(*string, *string) error) error {
return fmt.Errorf("%s, no such method", x.Name())
}
func (x *XinferenceModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig) (*OCRFileResponse, error) {
func (x *XinferenceModel) OCRFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, ocrConfig *OCRConfig, modelUsage *common.ModelUsage) (*OCRFileResponse, error) {
return nil, fmt.Errorf("%s, no such method", x.Name())
}
func (x *XinferenceModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig) (*ParseFileResponse, error) {
func (x *XinferenceModel) ParseFile(modelName *string, content []byte, url *string, apiConfig *APIConfig, parseFileConfig *ParseFileConfig, modelUsage *common.ModelUsage) (*ParseFileResponse, error) {
return nil, fmt.Errorf("%s, no such method", x.Name())
}