fix(tools): recover tool calls from tool_use_failed errors and update default space to digital-twin
This commit is contained in:
@@ -25,7 +25,7 @@ import (
|
||||
)
|
||||
|
||||
var (
|
||||
DefaultSpaceURL = "https://ghost2513-openai-gpt-oss-120b.hf.space"
|
||||
DefaultSpaceURL = "https://lucasmarchettidelima-digital-twin.hf.space"
|
||||
DefaultUserAgent = "Mozilla/5.0 (X11; Linux x86_64; rv:153.0) Gecko/20100101 Firefox/153.0"
|
||||
ConfiguredUserAgent string
|
||||
)
|
||||
@@ -388,6 +388,9 @@ func DoWithFibonacciRetry(client *http.Client, makeReq func() (*http.Request, er
|
||||
respBody, _ := io.ReadAll(resp.Body)
|
||||
resp.Body.Close()
|
||||
lastErr = fmt.Errorf("HTTP status %d: %s", resp.StatusCode, string(respBody))
|
||||
if _, ok := extractFailedGeneration(string(respBody)); ok {
|
||||
break
|
||||
}
|
||||
} else {
|
||||
lastErr = err
|
||||
}
|
||||
@@ -1002,6 +1005,141 @@ func WriteCompletionResponse(w http.ResponseWriter, completionID string, created
|
||||
json.NewEncoder(w).Encode(resp)
|
||||
}
|
||||
|
||||
func extractFailedGeneration(raw string) (string, bool) {
|
||||
var obj map[string]interface{}
|
||||
if err := json.Unmarshal([]byte(raw), &obj); err == nil {
|
||||
if fg, ok := obj["failed_generation"].(string); ok && fg != "" {
|
||||
return strings.TrimSpace(fg), true
|
||||
}
|
||||
if errVal, ok := obj["error"]; ok {
|
||||
if errMap, ok := errVal.(map[string]interface{}); ok {
|
||||
if fg, ok := errMap["failed_generation"].(string); ok && fg != "" {
|
||||
return strings.TrimSpace(fg), true
|
||||
}
|
||||
} else if errStr, ok := errVal.(string); ok {
|
||||
if fg, ok := extractFailedGeneration(errStr); ok {
|
||||
return fg, true
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
keyIdx := strings.Index(raw, `"failed_generation"`)
|
||||
if keyIdx == -1 {
|
||||
keyIdx = strings.Index(raw, `'failed_generation'`)
|
||||
}
|
||||
if keyIdx == -1 {
|
||||
keyIdx = strings.Index(raw, `failed_generation`)
|
||||
}
|
||||
if keyIdx == -1 {
|
||||
return "", false
|
||||
}
|
||||
|
||||
colonIdx := strings.Index(raw[keyIdx:], ":")
|
||||
if colonIdx == -1 {
|
||||
return "", false
|
||||
}
|
||||
valStart := keyIdx + colonIdx + 1
|
||||
|
||||
for valStart < len(raw) && (raw[valStart] == ' ' || raw[valStart] == '\t' || raw[valStart] == '\r' || raw[valStart] == '\n') {
|
||||
valStart++
|
||||
}
|
||||
if valStart >= len(raw) {
|
||||
return "", false
|
||||
}
|
||||
|
||||
firstChar := raw[valStart]
|
||||
var candidate string
|
||||
|
||||
if firstChar == '\'' || firstChar == '"' {
|
||||
quoteChar := firstChar
|
||||
var b strings.Builder
|
||||
escaped := false
|
||||
for i := valStart + 1; i < len(raw); i++ {
|
||||
ch := raw[i]
|
||||
if escaped {
|
||||
switch ch {
|
||||
case 'n':
|
||||
b.WriteByte('\n')
|
||||
case 'r':
|
||||
b.WriteByte('\r')
|
||||
case 't':
|
||||
b.WriteByte('\t')
|
||||
case '\\':
|
||||
b.WriteByte('\\')
|
||||
case '\'':
|
||||
b.WriteByte('\'')
|
||||
case '"':
|
||||
b.WriteByte('"')
|
||||
default:
|
||||
b.WriteByte('\\')
|
||||
b.WriteByte(ch)
|
||||
}
|
||||
escaped = false
|
||||
} else if ch == '\\' {
|
||||
escaped = true
|
||||
} else if ch == quoteChar {
|
||||
break
|
||||
} else {
|
||||
b.WriteByte(ch)
|
||||
}
|
||||
}
|
||||
candidate = strings.TrimSpace(b.String())
|
||||
} else if firstChar == '{' || firstChar == '[' {
|
||||
openChar := firstChar
|
||||
closeChar := byte('}')
|
||||
if openChar == '[' {
|
||||
closeChar = ']'
|
||||
}
|
||||
depth := 0
|
||||
inStr := false
|
||||
var strQuote byte
|
||||
escaped := false
|
||||
endIdx := -1
|
||||
|
||||
for i := valStart; i < len(raw); i++ {
|
||||
ch := raw[i]
|
||||
if inStr {
|
||||
if escaped {
|
||||
escaped = false
|
||||
} else if ch == '\\' {
|
||||
escaped = true
|
||||
} else if ch == strQuote {
|
||||
inStr = false
|
||||
}
|
||||
} else {
|
||||
if ch == '"' || ch == '\'' {
|
||||
inStr = true
|
||||
strQuote = ch
|
||||
} else if ch == openChar {
|
||||
depth++
|
||||
} else if ch == closeChar {
|
||||
depth--
|
||||
if depth == 0 {
|
||||
endIdx = i + 1
|
||||
break
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
if endIdx != -1 {
|
||||
candidate = strings.TrimSpace(raw[valStart:endIdx])
|
||||
}
|
||||
}
|
||||
|
||||
if candidate != "" {
|
||||
if strings.Contains(candidate, `\"`) {
|
||||
candidate = strings.ReplaceAll(candidate, `\"`, `"`)
|
||||
}
|
||||
if strings.Contains(candidate, `\n`) {
|
||||
candidate = strings.ReplaceAll(candidate, `\n`, "\n")
|
||||
}
|
||||
return candidate, true
|
||||
}
|
||||
|
||||
return "", false
|
||||
}
|
||||
|
||||
func extractGradioErrorMessage(dataStr string) string {
|
||||
var errObj map[string]interface{}
|
||||
if err := json.Unmarshal([]byte(dataStr), &errObj); err == nil {
|
||||
@@ -2339,6 +2477,9 @@ func (g *GradioGateway) ExecuteChatCompletion(w http.ResponseWriter, r *http.Req
|
||||
return fmt.Errorf("failed to encode request: %w", err)
|
||||
}
|
||||
|
||||
completionID := "chatcmpl-" + GenerateUUID()
|
||||
createdTime := time.Now().Unix()
|
||||
|
||||
// 1. Submit to /call/{endpoint}
|
||||
callURL := fmt.Sprintf("%s%s/call/%s", disc.SpaceURL, disc.APIPrefix, disc.CleanEndpoint)
|
||||
makeCallReq := func() (*http.Request, error) {
|
||||
@@ -2353,6 +2494,27 @@ func (g *GradioGateway) ExecuteChatCompletion(w http.ResponseWriter, r *http.Req
|
||||
|
||||
resp, err := DoWithFibonacciRetry(g.client, makeCallReq, 5)
|
||||
if err != nil {
|
||||
if fg, ok := extractFailedGeneration(err.Error()); ok {
|
||||
if tcs, _, has := DetectToolCalls(fg); has && len(tcs) > 0 {
|
||||
if !req.Stream {
|
||||
WriteCompletionResponse(w, completionID, createdTime, modelName, FinalOutput{
|
||||
ToolCalls: tcs,
|
||||
FinishReason: "tool_calls",
|
||||
})
|
||||
return nil
|
||||
}
|
||||
flusher, _ := w.(http.Flusher)
|
||||
streamer := NewStreamer(w, flusher, completionID, createdTime, modelName)
|
||||
for i, tc := range tcs {
|
||||
iCopy := i
|
||||
tc.Index = &iCopy
|
||||
streamer.ToolCallDelta(tc)
|
||||
}
|
||||
streamer.Finish("tool_calls")
|
||||
streamer.Done()
|
||||
return nil
|
||||
}
|
||||
}
|
||||
// If call failed, try without APIPrefix or try /call/v2
|
||||
altCallURL := fmt.Sprintf("%s/call/%s", disc.SpaceURL, disc.CleanEndpoint)
|
||||
makeAltReq := func() (*http.Request, error) {
|
||||
@@ -2366,6 +2528,27 @@ func (g *GradioGateway) ExecuteChatCompletion(w http.ResponseWriter, r *http.Req
|
||||
}
|
||||
resp, err = DoWithFibonacciRetry(g.client, makeAltReq, 3)
|
||||
if err != nil {
|
||||
if fg, ok := extractFailedGeneration(err.Error()); ok {
|
||||
if tcs, _, has := DetectToolCalls(fg); has && len(tcs) > 0 {
|
||||
if !req.Stream {
|
||||
WriteCompletionResponse(w, completionID, createdTime, modelName, FinalOutput{
|
||||
ToolCalls: tcs,
|
||||
FinishReason: "tool_calls",
|
||||
})
|
||||
return nil
|
||||
}
|
||||
flusher, _ := w.(http.Flusher)
|
||||
streamer := NewStreamer(w, flusher, completionID, createdTime, modelName)
|
||||
for i, tc := range tcs {
|
||||
iCopy := i
|
||||
tc.Index = &iCopy
|
||||
streamer.ToolCallDelta(tc)
|
||||
}
|
||||
streamer.Finish("tool_calls")
|
||||
streamer.Done()
|
||||
return nil
|
||||
}
|
||||
}
|
||||
return fmt.Errorf("upstream Gradio call error: %w", err)
|
||||
}
|
||||
}
|
||||
@@ -2390,13 +2573,31 @@ func (g *GradioGateway) ExecuteChatCompletion(w http.ResponseWriter, r *http.Req
|
||||
|
||||
streamResp, err := DoWithFibonacciRetry(g.client, makeStreamReq, 5)
|
||||
if err != nil {
|
||||
if fg, ok := extractFailedGeneration(err.Error()); ok {
|
||||
if tcs, _, has := DetectToolCalls(fg); has && len(tcs) > 0 {
|
||||
if !req.Stream {
|
||||
WriteCompletionResponse(w, completionID, createdTime, modelName, FinalOutput{
|
||||
ToolCalls: tcs,
|
||||
FinishReason: "tool_calls",
|
||||
})
|
||||
return nil
|
||||
}
|
||||
flusher, _ := w.(http.Flusher)
|
||||
streamer := NewStreamer(w, flusher, completionID, createdTime, modelName)
|
||||
for i, tc := range tcs {
|
||||
iCopy := i
|
||||
tc.Index = &iCopy
|
||||
streamer.ToolCallDelta(tc)
|
||||
}
|
||||
streamer.Finish("tool_calls")
|
||||
streamer.Done()
|
||||
return nil
|
||||
}
|
||||
}
|
||||
return fmt.Errorf("upstream Gradio stream error: %w", err)
|
||||
}
|
||||
defer streamResp.Body.Close()
|
||||
|
||||
completionID := "chatcmpl-" + GenerateUUID()
|
||||
createdTime := time.Now().Unix()
|
||||
|
||||
// 3. Handle Non-Streaming vs Streaming
|
||||
if !req.Stream {
|
||||
reader := bufio.NewReader(streamResp.Body)
|
||||
@@ -2418,6 +2619,15 @@ func (g *GradioGateway) ExecuteChatCompletion(w http.ResponseWriter, r *http.Req
|
||||
if strings.HasPrefix(line, "data: ") {
|
||||
dataStr := strings.TrimPrefix(line, "data: ")
|
||||
if currentEvent == "error" {
|
||||
if fg, ok := extractFailedGeneration(dataStr); ok {
|
||||
if tcs, _, has := DetectToolCalls(fg); has && len(tcs) > 0 {
|
||||
latestFrame = GradioOutputFrame{
|
||||
ToolCalls: tcs,
|
||||
OK: true,
|
||||
}
|
||||
break
|
||||
}
|
||||
}
|
||||
errMsg := extractGradioErrorMessage(dataStr)
|
||||
log.Printf("Upstream Gradio error: %s", errMsg)
|
||||
return fmt.Errorf("upstream Gradio error: %s", errMsg)
|
||||
@@ -2496,6 +2706,17 @@ func (g *GradioGateway) ExecuteChatCompletion(w http.ResponseWriter, r *http.Req
|
||||
if strings.HasPrefix(line, "data: ") {
|
||||
dataStr := strings.TrimPrefix(line, "data: ")
|
||||
if currentEvent == "error" {
|
||||
if fg, ok := extractFailedGeneration(dataStr); ok {
|
||||
if tcs, _, has := DetectToolCalls(fg); has && len(tcs) > 0 {
|
||||
for i, tc := range tcs {
|
||||
iCopy := i
|
||||
tc.Index = &iCopy
|
||||
streamer.ToolCallDelta(tc)
|
||||
}
|
||||
nativeToolCallsSeen = true
|
||||
break
|
||||
}
|
||||
}
|
||||
errMsg := extractGradioErrorMessage(dataStr)
|
||||
log.Printf("Upstream Gradio error: %s", errMsg)
|
||||
if !streamer.started {
|
||||
|
||||
Reference in New Issue
Block a user