fix(responses): include model and usage in translated streams

Ensure response.created and response.completed chunks produced by the OpenAI/Gemini/Claude translators always include required fields (response.model and response.usage) so clients validating Responses SSE do not fail schema validation.
2026-02-23 13:22:06 +08:00
parent 49c8ec69d0
commit 5382764d8a
6 changed files with 196 additions and 36 deletions
@@ -109,6 +109,7 @@ func ConvertClaudeResponseToOpenAIResponses(ctx context.Context, modelName strin
 			created, _ = sjson.Set(created, "sequence_number", nextSeq())
 			created, _ = sjson.Set(created, "response.id", st.ResponseID)
 			created, _ = sjson.Set(created, "response.created_at", st.CreatedAt)
+			created, _ = sjson.Set(created, "response.model", modelName)
 			out = append(out, emitEvent("response.created", created))
 			// response.in_progress
 			inprog := `{"type":"response.in_progress","sequence_number":0,"response":{"id":"","object":"response","created_at":0,"status":"in_progress"}}`
@@ -412,19 +413,14 @@ func ConvertClaudeResponseToOpenAIResponses(ctx context.Context, modelName strin
 		if st.ReasoningBuf.Len() > 0 {
 			reasoningTokens = int64(st.ReasoningBuf.Len() / 4)
 		}
-		usagePresent := st.UsageSeen || reasoningTokens > 0
-		if usagePresent {
-			completed, _ = sjson.Set(completed, "response.usage.input_tokens", st.InputTokens)
-			completed, _ = sjson.Set(completed, "response.usage.input_tokens_details.cached_tokens", 0)
-			completed, _ = sjson.Set(completed, "response.usage.output_tokens", st.OutputTokens)
-			if reasoningTokens > 0 {
-				completed, _ = sjson.Set(completed, "response.usage.output_tokens_details.reasoning_tokens", reasoningTokens)
-			}
-			total := st.InputTokens + st.OutputTokens
-			if total > 0 || st.UsageSeen {
-				completed, _ = sjson.Set(completed, "response.usage.total_tokens", total)
-			}
+		completed, _ = sjson.Set(completed, "response.usage.input_tokens", st.InputTokens)
+		completed, _ = sjson.Set(completed, "response.usage.input_tokens_details.cached_tokens", 0)
+		completed, _ = sjson.Set(completed, "response.usage.output_tokens", st.OutputTokens)
+		if reasoningTokens > 0 {
+			completed, _ = sjson.Set(completed, "response.usage.output_tokens_details.reasoning_tokens", reasoningTokens)
 		}
+		total := st.InputTokens + st.OutputTokens
+		completed, _ = sjson.Set(completed, "response.usage.total_tokens", total)
 		out = append(out, emitEvent("response.completed", completed))
 	}

@@ -0,0 +1,67 @@
+package responses
+
+import (
+	"context"
+	"strings"
+	"testing"
+
+	"github.com/tidwall/gjson"
+)
+
+func parseSSEEvent(t *testing.T, chunk string) (string, gjson.Result) {
+	t.Helper()
+
+	lines := strings.Split(chunk, "\n")
+	if len(lines) < 2 {
+		t.Fatalf("unexpected SSE chunk: %q", chunk)
+	}
+
+	event := strings.TrimSpace(strings.TrimPrefix(lines[0], "event:"))
+	dataLine := strings.TrimSpace(strings.TrimPrefix(lines[1], "data:"))
+	if !gjson.Valid(dataLine) {
+		t.Fatalf("invalid SSE data JSON: %q", dataLine)
+	}
+	return event, gjson.Parse(dataLine)
+}
+
+func TestConvertClaudeResponseToOpenAIResponses_CreatedHasModelAndCompletedHasUsage(t *testing.T) {
+	in := []string{
+		`data: {"type":"message_start","message":{"id":"msg_1"}}`,
+		`data: {"type":"message_stop"}`,
+	}
+
+	var param any
+	var out []string
+	for _, line := range in {
+		out = append(out, ConvertClaudeResponseToOpenAIResponses(context.Background(), "test-model", nil, nil, []byte(line), &param)...)
+	}
+
+	gotCreated := false
+	gotCompleted := false
+	createdModel := ""
+	for _, chunk := range out {
+		ev, data := parseSSEEvent(t, chunk)
+		switch ev {
+		case "response.created":
+			gotCreated = true
+			createdModel = data.Get("response.model").String()
+		case "response.completed":
+			gotCompleted = true
+			if !data.Get("response.usage.input_tokens").Exists() {
+				t.Fatalf("response.completed missing usage.input_tokens: %s", data.Raw)
+			}
+			if !data.Get("response.usage.output_tokens").Exists() {
+				t.Fatalf("response.completed missing usage.output_tokens: %s", data.Raw)
+			}
+		}
+	}
+	if !gotCreated {
+		t.Fatalf("missing response.created event")
+	}
+	if createdModel != "test-model" {
+		t.Fatalf("unexpected response.created model: got %q", createdModel)
+	}
+	if !gotCompleted {
+		t.Fatalf("missing response.completed event")
+	}
+}