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.

internal/translator/claude/openai/responses/claude_openai-responses_response.go

@@ -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"}}`
@@ -412,8 +413,6 @@ 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)
@@ -421,10 +420,7 @@ func ConvertClaudeResponseToOpenAIResponses(ctx context.Context, modelName strin
 			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)
-			}
-		}
 		out = append(out, emitEvent("response.completed", completed))
 	}
 

internal/translator/claude/openai/responses/claude_openai-responses_response_test.go

@@ -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")
+	}
+}

internal/translator/gemini/openai/responses/gemini_openai-responses_response.go

@@ -212,6 +212,7 @@ func ConvertGeminiResponseToOpenAIResponses(_ context.Context, modelName string,
 		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))
 
 		inprog := `{"type":"response.in_progress","sequence_number":0,"response":{"id":"","object":"response","created_at":0,"status":"in_progress"}}`
@@ -529,31 +530,36 @@ func ConvertGeminiResponseToOpenAIResponses(_ context.Context, modelName string,
 			completed, _ = sjson.SetRaw(completed, "response.output", gjson.Get(outputsWrapper, "arr").Raw)
 		}
 
-		// usage mapping
+		input := int64(0)
+		cached := int64(0)
+		output := int64(0)
+		reasoning := int64(0)
+		total := int64(0)
 		if um := root.Get("usageMetadata"); um.Exists() {
 			// input tokens = prompt + thoughts
-			input := um.Get("promptTokenCount").Int() + um.Get("thoughtsTokenCount").Int()
+			input = um.Get("promptTokenCount").Int() + um.Get("thoughtsTokenCount").Int()
-			completed, _ = sjson.Set(completed, "response.usage.input_tokens", input)
 			// cached token details: align with OpenAI "cached_tokens" semantics.
-			completed, _ = sjson.Set(completed, "response.usage.input_tokens_details.cached_tokens", um.Get("cachedContentTokenCount").Int())
+			cached = um.Get("cachedContentTokenCount").Int()
 			// output tokens
 			if v := um.Get("candidatesTokenCount"); v.Exists() {
-				completed, _ = sjson.Set(completed, "response.usage.output_tokens", v.Int())
+				output = v.Int()
-			} else {
-				completed, _ = sjson.Set(completed, "response.usage.output_tokens", 0)
 			}
 			if v := um.Get("thoughtsTokenCount"); v.Exists() {
-				completed, _ = sjson.Set(completed, "response.usage.output_tokens_details.reasoning_tokens", v.Int())
+				reasoning = v.Int()
-			} else {
-				completed, _ = sjson.Set(completed, "response.usage.output_tokens_details.reasoning_tokens", 0)
 			}
 			if v := um.Get("totalTokenCount"); v.Exists() {
-				completed, _ = sjson.Set(completed, "response.usage.total_tokens", v.Int())
+				total = v.Int()
 			} else {
-				completed, _ = sjson.Set(completed, "response.usage.total_tokens", 0)
+				total = input + output
 			}
 		}
 
+		completed, _ = sjson.Set(completed, "response.usage.input_tokens", input)
+		completed, _ = sjson.Set(completed, "response.usage.input_tokens_details.cached_tokens", cached)
+		completed, _ = sjson.Set(completed, "response.usage.output_tokens", output)
+		completed, _ = sjson.Set(completed, "response.usage.output_tokens_details.reasoning_tokens", reasoning)
+		completed, _ = sjson.Set(completed, "response.usage.total_tokens", total)
+
 		out = append(out, emitEvent("response.completed", completed))
 	}
 

internal/translator/gemini/openai/responses/gemini_openai-responses_response_test.go

@@ -53,6 +53,7 @@ func TestConvertGeminiResponseToOpenAIResponses_UnwrapAndAggregateText(t *testin
 		textDone     string
 		messageText  string
 		responseID   string
+		createdModel string
 		instructions string
 		cachedTokens int64
 
@@ -68,6 +69,8 @@ func TestConvertGeminiResponseToOpenAIResponses_UnwrapAndAggregateText(t *testin
 	for i, chunk := range out {
 		ev, data := parseSSEEvent(t, chunk)
 		switch ev {
+		case "response.created":
+			createdModel = data.Get("response.model").String()
 		case "response.output_text.done":
 			gotTextDone = true
 			if posTextDone == -1 {
@@ -132,6 +135,9 @@ func TestConvertGeminiResponseToOpenAIResponses_UnwrapAndAggregateText(t *testin
 	if responseID != "resp_req_vrtx_1" {
 		t.Fatalf("unexpected response id: got %q", responseID)
 	}
+	if createdModel != "test-model" {
+		t.Fatalf("unexpected response.created model: got %q", createdModel)
+	}
 	if instructions != "test instructions" {
 		t.Fatalf("unexpected instructions echo: got %q", instructions)
 	}
@@ -153,6 +159,31 @@ func TestConvertGeminiResponseToOpenAIResponses_UnwrapAndAggregateText(t *testin
 	}
 }
 
+func TestConvertGeminiResponseToOpenAIResponses_CompletedAlwaysHasUsage(t *testing.T) {
+	in := `data: {"response":{"candidates":[{"content":{"role":"model","parts":[{"text":"hi"}]},"finishReason":"STOP"}],"modelVersion":"test-model","responseId":"req_no_usage"},"traceId":"t1"}`
+
+	var param any
+	out := ConvertGeminiResponseToOpenAIResponses(context.Background(), "test-model", nil, nil, []byte(in), &param)
+
+	gotCompleted := false
+	for _, chunk := range out {
+		ev, data := parseSSEEvent(t, chunk)
+		if ev != "response.completed" {
+			continue
+		}
+		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 !gotCompleted {
+		t.Fatalf("missing response.completed event")
+	}
+}
+
 func TestConvertGeminiResponseToOpenAIResponses_ReasoningEncryptedContent(t *testing.T) {
 	sig := "RXE0RENrZ0lDeEFDR0FJcVFOZDdjUzlleGFuRktRdFcvSzNyZ2MvWDNCcDQ4RmxSbGxOWUlOVU5kR1l1UHMrMGdkMVp0Vkg3ekdKU0g4YVljc2JjN3lNK0FrdGpTNUdqamI4T3Z0VVNETzdQd3pmcFhUOGl3U3hXUEJvTVFRQ09mWTFyMEtTWGZxUUlJakFqdmFGWk83RW1XRlBKckJVOVpkYzdDKw=="
 	in := []string{

internal/translator/openai/openai/responses/openai_openai-responses_response.go

@@ -153,6 +153,7 @@ func ConvertOpenAIChatCompletionsResponseToOpenAIResponses(ctx context.Context,
 		created, _ = sjson.Set(created, "sequence_number", nextSeq())
 		created, _ = sjson.Set(created, "response.id", st.ResponseID)
 		created, _ = sjson.Set(created, "response.created_at", st.Created)
+		created, _ = sjson.Set(created, "response.model", modelName)
 		out = append(out, emitRespEvent("response.created", created))
 
 		inprog := `{"type":"response.in_progress","sequence_number":0,"response":{"id":"","object":"response","created_at":0,"status":"in_progress"}}`
@@ -578,7 +579,6 @@ func ConvertOpenAIChatCompletionsResponseToOpenAIResponses(ctx context.Context,
 				if gjson.Get(outputsWrapper, "arr.#").Int() > 0 {
 					completed, _ = sjson.SetRaw(completed, "response.output", gjson.Get(outputsWrapper, "arr").Raw)
 				}
-				if st.UsageSeen {
 				completed, _ = sjson.Set(completed, "response.usage.input_tokens", st.PromptTokens)
 				completed, _ = sjson.Set(completed, "response.usage.input_tokens_details.cached_tokens", st.CachedTokens)
 				completed, _ = sjson.Set(completed, "response.usage.output_tokens", st.CompletionTokens)
@@ -590,7 +590,6 @@ func ConvertOpenAIChatCompletionsResponseToOpenAIResponses(ctx context.Context,
 					total = st.PromptTokens + st.CompletionTokens
 				}
 				completed, _ = sjson.Set(completed, "response.usage.total_tokens", total)
-				}
 				out = append(out, emitRespEvent("response.completed", completed))
 			}
 

internal/translator/openai/openai/responses/openai_openai-responses_response_test.go

@@ -0,0 +1,61 @@
+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 TestConvertOpenAIChatCompletionsResponseToOpenAIResponses_CreatedHasModelAndCompletedHasUsage(t *testing.T) {
+	in := `data: {"id":"chatcmpl-1","object":"chat.completion.chunk","created":1700000000,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}`
+
+	var param any
+	out := ConvertOpenAIChatCompletionsResponseToOpenAIResponses(context.Background(), "test-model", nil, nil, []byte(in), &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")
+	}
+}