diff --git a/internal/models/chat/remote_api.go b/internal/models/chat/remote_api.go index ab16f794..9b327a0e 100644 --- a/internal/models/chat/remote_api.go +++ b/internal/models/chat/remote_api.go @@ -546,7 +546,20 @@ func (c *RemoteAPIChat) processRawHTTPStream(ctx context.Context, resp *http.Res for { event, err := reader.ReadEvent() if err != nil { - if err != io.EOF { + if err == io.EOF { + // 部分模型不发送 [DONE] 标记,直接关闭连接,视为正常结束 + if state.usage != nil { + logger.Infof(ctx, "[LLM Usage] model=%s, prompt_tokens=%d, completion_tokens=%d, total_tokens=%d", + c.modelName, state.usage.PromptTokens, state.usage.CompletionTokens, state.usage.TotalTokens) + } + streamChan <- types.StreamResponse{ + ResponseType: types.ResponseTypeAnswer, + Content: "", + Done: true, + ToolCalls: state.buildOrderedToolCalls(), + Usage: state.usage, + } + } else { logger.Errorf(ctx, "Stream read error: %v", err) streamChan <- types.StreamResponse{ ResponseType: types.ResponseTypeError, diff --git a/internal/models/chat/sse_reader.go b/internal/models/chat/sse_reader.go index 45c6cf2b..6877be27 100644 --- a/internal/models/chat/sse_reader.go +++ b/internal/models/chat/sse_reader.go @@ -2,7 +2,6 @@ package chat import ( "bufio" - "errors" "io" "strings" ) @@ -55,5 +54,5 @@ func (r *SSEReader) ReadEvent() (*SSEEvent, error) { return nil, err } - return nil, errors.New("EOF") + return nil, io.EOF }