Add OpenAI Responses support

This commit is contained in:
Luis Pater
2025-09-01 17:38:24 +08:00
parent e11637dc62
commit 3acdec51bd
54 changed files with 2394 additions and 144 deletions

View File

@@ -181,6 +181,7 @@ func (c *ClaudeClient) TokenStorage() auth.TokenStorage {
// - []byte: The response body.
// - *interfaces.ErrorMessage: An error message if the request fails.
func (c *ClaudeClient) SendRawMessage(ctx context.Context, modelName string, rawJSON []byte, alt string) ([]byte, *interfaces.ErrorMessage) {
originalRequestRawJSON := bytes.Clone(rawJSON)
handler := ctx.Value("handler").(interfaces.APIHandler)
handlerType := handler.HandlerType()
rawJSON = translator.Request(handlerType, c.Type(), modelName, rawJSON, false)
@@ -208,7 +209,7 @@ func (c *ClaudeClient) SendRawMessage(ctx context.Context, modelName string, raw
c.AddAPIResponseData(ctx, bodyBytes)
var param any
bodyBytes = []byte(translator.ResponseNonStream(handlerType, c.Type(), ctx, modelName, bodyBytes, &param))
bodyBytes = []byte(translator.ResponseNonStream(handlerType, c.Type(), ctx, modelName, originalRequestRawJSON, rawJSON, bodyBytes, &param))
return bodyBytes, nil
}
@@ -226,6 +227,8 @@ func (c *ClaudeClient) SendRawMessage(ctx context.Context, modelName string, raw
// - <-chan []byte: A channel for receiving response data chunks.
// - <-chan *interfaces.ErrorMessage: A channel for receiving error messages.
func (c *ClaudeClient) SendRawMessageStream(ctx context.Context, modelName string, rawJSON []byte, alt string) (<-chan []byte, <-chan *interfaces.ErrorMessage) {
originalRequestRawJSON := bytes.Clone(rawJSON)
handler := ctx.Value("handler").(interfaces.APIHandler)
handlerType := handler.HandlerType()
rawJSON = translator.Request(handlerType, c.Type(), modelName, rawJSON, true)
@@ -275,7 +278,7 @@ func (c *ClaudeClient) SendRawMessageStream(ctx context.Context, modelName strin
var param any
for scanner.Scan() {
line := scanner.Bytes()
lines := translator.Response(handlerType, c.Type(), ctx, modelName, line, &param)
lines := translator.Response(handlerType, c.Type(), ctx, modelName, originalRequestRawJSON, rawJSON, line, &param)
for i := 0; i < len(lines); i++ {
dataChan <- []byte(lines[i])
}

View File

@@ -124,6 +124,8 @@ func (c *CodexClient) TokenStorage() auth.TokenStorage {
// - []byte: The response body.
// - *interfaces.ErrorMessage: An error message if the request fails.
func (c *CodexClient) SendRawMessage(ctx context.Context, modelName string, rawJSON []byte, alt string) ([]byte, *interfaces.ErrorMessage) {
originalRequestRawJSON := bytes.Clone(rawJSON)
handler := ctx.Value("handler").(interfaces.APIHandler)
handlerType := handler.HandlerType()
rawJSON = translator.Request(handlerType, c.Type(), modelName, rawJSON, false)
@@ -150,7 +152,7 @@ func (c *CodexClient) SendRawMessage(ctx context.Context, modelName string, rawJ
c.AddAPIResponseData(ctx, bodyBytes)
var param any
bodyBytes = []byte(translator.ResponseNonStream(handlerType, c.Type(), ctx, modelName, bodyBytes, &param))
bodyBytes = []byte(translator.ResponseNonStream(handlerType, c.Type(), ctx, modelName, originalRequestRawJSON, rawJSON, bodyBytes, &param))
return bodyBytes, nil
@@ -168,6 +170,8 @@ func (c *CodexClient) SendRawMessage(ctx context.Context, modelName string, rawJ
// - <-chan []byte: A channel for receiving response data chunks.
// - <-chan *interfaces.ErrorMessage: A channel for receiving error messages.
func (c *CodexClient) SendRawMessageStream(ctx context.Context, modelName string, rawJSON []byte, alt string) (<-chan []byte, <-chan *interfaces.ErrorMessage) {
originalRequestRawJSON := bytes.Clone(rawJSON)
handler := ctx.Value("handler").(interfaces.APIHandler)
handlerType := handler.HandlerType()
rawJSON = translator.Request(handlerType, c.Type(), modelName, rawJSON, true)
@@ -218,7 +222,7 @@ func (c *CodexClient) SendRawMessageStream(ctx context.Context, modelName string
var param any
for scanner.Scan() {
line := scanner.Bytes()
lines := translator.Response(handlerType, c.Type(), ctx, modelName, line, &param)
lines := translator.Response(handlerType, c.Type(), ctx, modelName, originalRequestRawJSON, rawJSON, line, &param)
for i := 0; i < len(lines); i++ {
dataChan <- []byte(lines[i])
}

View File

@@ -407,6 +407,7 @@ func (c *GeminiCLIClient) APIRequest(ctx context.Context, modelName, endpoint st
// - []byte: The response body.
// - *interfaces.ErrorMessage: An error message if the request fails.
func (c *GeminiCLIClient) SendRawTokenCount(ctx context.Context, modelName string, rawJSON []byte, alt string) ([]byte, *interfaces.ErrorMessage) {
originalRequestRawJSON := bytes.Clone(rawJSON)
for {
if c.isModelQuotaExceeded(modelName) {
if c.cfg.QuotaExceeded.SwitchPreviewModel {
@@ -453,7 +454,7 @@ func (c *GeminiCLIClient) SendRawTokenCount(ctx context.Context, modelName strin
c.AddAPIResponseData(ctx, bodyBytes)
var param any
bodyBytes = []byte(translator.ResponseNonStream(handlerType, c.Type(), ctx, modelName, bodyBytes, &param))
bodyBytes = []byte(translator.ResponseNonStream(handlerType, c.Type(), ctx, modelName, originalRequestRawJSON, rawJSON, bodyBytes, &param))
return bodyBytes, nil
}
@@ -471,6 +472,8 @@ func (c *GeminiCLIClient) SendRawTokenCount(ctx context.Context, modelName strin
// - []byte: The response body.
// - *interfaces.ErrorMessage: An error message if the request fails.
func (c *GeminiCLIClient) SendRawMessage(ctx context.Context, modelName string, rawJSON []byte, alt string) ([]byte, *interfaces.ErrorMessage) {
originalRequestRawJSON := bytes.Clone(rawJSON)
handler := ctx.Value("handler").(interfaces.APIHandler)
handlerType := handler.HandlerType()
rawJSON = translator.Request(handlerType, c.Type(), modelName, rawJSON, false)
@@ -484,6 +487,7 @@ func (c *GeminiCLIClient) SendRawMessage(ctx context.Context, modelName string,
if newModelName != "" {
log.Debugf("Model %s is quota exceeded. Switch to preview model %s", modelName, newModelName)
rawJSON, _ = sjson.SetBytes(rawJSON, "model", newModelName)
modelName = newModelName
continue
}
}
@@ -519,7 +523,7 @@ func (c *GeminiCLIClient) SendRawMessage(ctx context.Context, modelName string,
newCtx := context.WithValue(ctx, "alt", alt)
var param any
bodyBytes = []byte(translator.ResponseNonStream(handlerType, c.Type(), newCtx, modelName, bodyBytes, &param))
bodyBytes = []byte(translator.ResponseNonStream(handlerType, c.Type(), newCtx, modelName, originalRequestRawJSON, rawJSON, bodyBytes, &param))
return bodyBytes, nil
}
@@ -537,6 +541,8 @@ func (c *GeminiCLIClient) SendRawMessage(ctx context.Context, modelName string,
// - <-chan []byte: A channel for receiving response data chunks.
// - <-chan *interfaces.ErrorMessage: A channel for receiving error messages.
func (c *GeminiCLIClient) SendRawMessageStream(ctx context.Context, modelName string, rawJSON []byte, alt string) (<-chan []byte, <-chan *interfaces.ErrorMessage) {
originalRequestRawJSON := bytes.Clone(rawJSON)
handler := ctx.Value("handler").(interfaces.APIHandler)
handlerType := handler.HandlerType()
rawJSON = translator.Request(handlerType, c.Type(), modelName, rawJSON, true)
@@ -563,6 +569,7 @@ func (c *GeminiCLIClient) SendRawMessageStream(ctx context.Context, modelName st
if newModelName != "" {
log.Debugf("Model %s is quota exceeded. Switch to preview model %s", modelName, newModelName)
rawJSON, _ = sjson.SetBytes(rawJSON, "model", newModelName)
modelName = newModelName
continue
}
}
@@ -608,7 +615,7 @@ func (c *GeminiCLIClient) SendRawMessageStream(ctx context.Context, modelName st
for scanner.Scan() {
line := scanner.Bytes()
if bytes.HasPrefix(line, dataTag) {
lines := translator.Response(handlerType, c.Type(), newCtx, modelName, line[6:], &param)
lines := translator.Response(handlerType, c.Type(), newCtx, modelName, originalRequestRawJSON, rawJSON, line[6:], &param)
for i := 0; i < len(lines); i++ {
dataChan <- []byte(lines[i])
}
@@ -640,7 +647,7 @@ func (c *GeminiCLIClient) SendRawMessageStream(ctx context.Context, modelName st
}
if translator.NeedConvert(handlerType, c.Type()) {
lines := translator.Response(handlerType, c.Type(), newCtx, modelName, data, &param)
lines := translator.Response(handlerType, c.Type(), newCtx, modelName, originalRequestRawJSON, rawJSON, data, &param)
for i := 0; i < len(lines); i++ {
dataChan <- []byte(lines[i])
}
@@ -651,7 +658,7 @@ func (c *GeminiCLIClient) SendRawMessageStream(ctx context.Context, modelName st
}
if translator.NeedConvert(handlerType, c.Type()) {
lines := translator.Response(handlerType, c.Type(), ctx, modelName, []byte("[DONE]"), &param)
lines := translator.Response(handlerType, c.Type(), ctx, modelName, rawJSON, originalRequestRawJSON, []byte("[DONE]"), &param)
for i := 0; i < len(lines); i++ {
dataChan <- []byte(lines[i])
}

View File

@@ -187,6 +187,7 @@ func (c *GeminiClient) APIRequest(ctx context.Context, modelName, endpoint strin
// - []byte: The response body.
// - *interfaces.ErrorMessage: An error message if the request fails.
func (c *GeminiClient) SendRawTokenCount(ctx context.Context, modelName string, rawJSON []byte, alt string) ([]byte, *interfaces.ErrorMessage) {
originalRequestRawJSON := bytes.Clone(rawJSON)
for {
if c.IsModelQuotaExceeded(modelName) {
return nil, &interfaces.ErrorMessage{
@@ -219,7 +220,7 @@ func (c *GeminiClient) SendRawTokenCount(ctx context.Context, modelName string,
c.AddAPIResponseData(ctx, bodyBytes)
var param any
bodyBytes = []byte(translator.ResponseNonStream(handlerType, c.Type(), ctx, modelName, bodyBytes, &param))
bodyBytes = []byte(translator.ResponseNonStream(handlerType, c.Type(), ctx, modelName, originalRequestRawJSON, rawJSON, bodyBytes, &param))
return bodyBytes, nil
}
@@ -237,6 +238,8 @@ func (c *GeminiClient) SendRawTokenCount(ctx context.Context, modelName string,
// - []byte: The response body.
// - *interfaces.ErrorMessage: An error message if the request fails.
func (c *GeminiClient) SendRawMessage(ctx context.Context, modelName string, rawJSON []byte, alt string) ([]byte, *interfaces.ErrorMessage) {
originalRequestRawJSON := bytes.Clone(rawJSON)
handler := ctx.Value("handler").(interfaces.APIHandler)
handlerType := handler.HandlerType()
rawJSON = translator.Request(handlerType, c.Type(), modelName, rawJSON, false)
@@ -268,11 +271,12 @@ func (c *GeminiClient) SendRawMessage(ctx context.Context, modelName string, raw
_ = respBody.Close()
c.AddAPIResponseData(ctx, bodyBytes)
// log.Debugf("Gemini response: %s", string(bodyBytes))
var param any
bodyBytes = []byte(translator.ResponseNonStream(handlerType, c.Type(), ctx, modelName, bodyBytes, &param))
output := []byte(translator.ResponseNonStream(handlerType, c.Type(), ctx, modelName, originalRequestRawJSON, rawJSON, bodyBytes, &param))
return bodyBytes, nil
return output, nil
}
// SendRawMessageStream handles a single conversational turn, including tool calls.
@@ -287,6 +291,8 @@ func (c *GeminiClient) SendRawMessage(ctx context.Context, modelName string, raw
// - <-chan []byte: A channel for receiving response data chunks.
// - <-chan *interfaces.ErrorMessage: A channel for receiving error messages.
func (c *GeminiClient) SendRawMessageStream(ctx context.Context, modelName string, rawJSON []byte, alt string) (<-chan []byte, <-chan *interfaces.ErrorMessage) {
originalRequestRawJSON := bytes.Clone(rawJSON)
handler := ctx.Value("handler").(interfaces.APIHandler)
handlerType := handler.HandlerType()
rawJSON = translator.Request(handlerType, c.Type(), modelName, rawJSON, true)
@@ -335,7 +341,7 @@ func (c *GeminiClient) SendRawMessageStream(ctx context.Context, modelName strin
for scanner.Scan() {
line := scanner.Bytes()
if bytes.HasPrefix(line, dataTag) {
lines := translator.Response(handlerType, c.Type(), newCtx, modelName, line[6:], &param)
lines := translator.Response(handlerType, c.Type(), newCtx, modelName, originalRequestRawJSON, rawJSON, line[6:], &param)
for i := 0; i < len(lines); i++ {
dataChan <- []byte(lines[i])
}
@@ -367,7 +373,7 @@ func (c *GeminiClient) SendRawMessageStream(ctx context.Context, modelName strin
}
if translator.NeedConvert(handlerType, c.Type()) {
lines := translator.Response(handlerType, c.Type(), newCtx, modelName, data, &param)
lines := translator.Response(handlerType, c.Type(), newCtx, modelName, originalRequestRawJSON, rawJSON, data, &param)
for i := 0; i < len(lines); i++ {
dataChan <- []byte(lines[i])
}
@@ -379,7 +385,7 @@ func (c *GeminiClient) SendRawMessageStream(ctx context.Context, modelName strin
}
if translator.NeedConvert(handlerType, c.Type()) {
lines := translator.Response(handlerType, c.Type(), ctx, modelName, []byte("[DONE]"), &param)
lines := translator.Response(handlerType, c.Type(), ctx, modelName, rawJSON, originalRequestRawJSON, []byte("[DONE]"), &param)
for i := 0; i < len(lines); i++ {
dataChan <- []byte(lines[i])
}

View File

@@ -199,6 +199,12 @@ func (c *OpenAICompatibilityClient) APIRequest(ctx context.Context, modelName st
log.Debugf("OpenAI Compatibility [%s] API request: %s", c.compatConfig.Name, util.HideAPIKey(apiKey))
if c.cfg.RequestLog {
if ginContext, ok := ctx.Value("gin").(*gin.Context); ok {
ginContext.Set("API_REQUEST", modifiedJSON)
}
}
// Send the request
resp, err := c.httpClient.Do(req)
if err != nil {
@@ -231,6 +237,8 @@ func (c *OpenAICompatibilityClient) APIRequest(ctx context.Context, modelName st
// - []byte: The response data from the API.
// - *interfaces.ErrorMessage: An error message if the request fails.
func (c *OpenAICompatibilityClient) SendRawMessage(ctx context.Context, modelName string, rawJSON []byte, alt string) ([]byte, *interfaces.ErrorMessage) {
originalRequestRawJSON := bytes.Clone(rawJSON)
handler := ctx.Value("handler").(interfaces.APIHandler)
handlerType := handler.HandlerType()
rawJSON = translator.Request(handlerType, c.Type(), modelName, rawJSON, false)
@@ -257,7 +265,7 @@ func (c *OpenAICompatibilityClient) SendRawMessage(ctx context.Context, modelNam
c.AddAPIResponseData(ctx, bodyBytes)
var param any
bodyBytes = []byte(translator.ResponseNonStream(handlerType, c.Type(), ctx, modelName, bodyBytes, &param))
bodyBytes = []byte(translator.ResponseNonStream(handlerType, c.Type(), ctx, modelName, originalRequestRawJSON, rawJSON, bodyBytes, &param))
return bodyBytes, nil
}
@@ -274,11 +282,14 @@ func (c *OpenAICompatibilityClient) SendRawMessage(ctx context.Context, modelNam
// - <-chan []byte: A channel that will receive response chunks.
// - <-chan *interfaces.ErrorMessage: A channel that will receive error messages.
func (c *OpenAICompatibilityClient) SendRawMessageStream(ctx context.Context, modelName string, rawJSON []byte, alt string) (<-chan []byte, <-chan *interfaces.ErrorMessage) {
originalRequestRawJSON := bytes.Clone(rawJSON)
handler := ctx.Value("handler").(interfaces.APIHandler)
handlerType := handler.HandlerType()
rawJSON = translator.Request(handlerType, c.Type(), modelName, rawJSON, true)
dataTag := []byte("data: ")
dataUglyTag := []byte("data:") // Some APIs providers don't add space after "data:", fuck for them all
doneTag := []byte("data: [DONE]")
errChan := make(chan *interfaces.ErrorMessage)
dataChan := make(chan []byte)
@@ -321,7 +332,33 @@ func (c *OpenAICompatibilityClient) SendRawMessageStream(ctx context.Context, mo
if bytes.Equal(line, doneTag) {
break
}
lines := translator.Response(handlerType, c.Type(), newCtx, modelName, line[6:], &param)
lines := translator.Response(handlerType, c.Type(), newCtx, modelName, originalRequestRawJSON, rawJSON, line[6:], &param)
for i := 0; i < len(lines); i++ {
c.AddAPIResponseData(ctx, line)
dataChan <- []byte(lines[i])
}
} else if bytes.HasPrefix(line, dataUglyTag) {
if bytes.Equal(line, doneTag) {
break
}
lines := translator.Response(handlerType, c.Type(), newCtx, modelName, originalRequestRawJSON, rawJSON, line[5:], &param)
for i := 0; i < len(lines); i++ {
c.AddAPIResponseData(ctx, line)
dataChan <- []byte(lines[i])
}
} else if bytes.HasPrefix(line, dataUglyTag) {
if bytes.Equal(line, doneTag) {
break
}
lines := translator.Response(handlerType, c.Type(), newCtx, modelName, line[5:], &param)
for i := 0; i < len(lines); i++ {
dataChan <- []byte(lines[i])
}
} else if bytes.HasPrefix(line, dataUglyTag) {
if bytes.Equal(line, doneTag) {
break
}
lines := translator.Response(handlerType, c.Type(), newCtx, modelName, line[5:], &param)
for i := 0; i < len(lines); i++ {
dataChan <- []byte(lines[i])
}
@@ -337,6 +374,9 @@ func (c *OpenAICompatibilityClient) SendRawMessageStream(ctx context.Context, mo
}
c.AddAPIResponseData(newCtx, line[6:])
dataChan <- line[6:]
} else if bytes.HasPrefix(line, dataUglyTag) {
c.AddAPIResponseData(newCtx, line[5:])
dataChan <- line[5:]
}
}
}

View File

@@ -119,6 +119,8 @@ func (c *QwenClient) TokenStorage() auth.TokenStorage {
// - []byte: The response body.
// - *interfaces.ErrorMessage: An error message if the request fails.
func (c *QwenClient) SendRawMessage(ctx context.Context, modelName string, rawJSON []byte, alt string) ([]byte, *interfaces.ErrorMessage) {
originalRequestRawJSON := bytes.Clone(rawJSON)
handler := ctx.Value("handler").(interfaces.APIHandler)
handlerType := handler.HandlerType()
rawJSON = translator.Request(handlerType, c.Type(), modelName, rawJSON, false)
@@ -145,7 +147,7 @@ func (c *QwenClient) SendRawMessage(ctx context.Context, modelName string, rawJS
c.AddAPIResponseData(ctx, bodyBytes)
var param any
bodyBytes = []byte(translator.ResponseNonStream(handlerType, c.Type(), ctx, modelName, bodyBytes, &param))
bodyBytes = []byte(translator.ResponseNonStream(handlerType, c.Type(), ctx, modelName, originalRequestRawJSON, rawJSON, bodyBytes, &param))
return bodyBytes, nil
@@ -163,6 +165,8 @@ func (c *QwenClient) SendRawMessage(ctx context.Context, modelName string, rawJS
// - <-chan []byte: A channel for receiving response data chunks.
// - <-chan *interfaces.ErrorMessage: A channel for receiving error messages.
func (c *QwenClient) SendRawMessageStream(ctx context.Context, modelName string, rawJSON []byte, alt string) (<-chan []byte, <-chan *interfaces.ErrorMessage) {
originalRequestRawJSON := bytes.Clone(rawJSON)
handler := ctx.Value("handler").(interfaces.APIHandler)
handlerType := handler.HandlerType()
rawJSON = translator.Request(handlerType, c.Type(), modelName, rawJSON, true)
@@ -216,7 +220,7 @@ func (c *QwenClient) SendRawMessageStream(ctx context.Context, modelName string,
for scanner.Scan() {
line := scanner.Bytes()
if bytes.HasPrefix(line, dataTag) {
lines := translator.Response(handlerType, c.Type(), ctx, modelName, line[6:], &param)
lines := translator.Response(handlerType, c.Type(), ctx, modelName, originalRequestRawJSON, rawJSON, line[6:], &param)
for i := 0; i < len(lines); i++ {
dataChan <- []byte(lines[i])
}