fix(service): preserve anthropic usage fields across compat endpoints

Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent) Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
2026-03-24 09:32:34 +08:00
parent 5418e15e63
commit 2f8e10db46
4 changed files with 317 additions and 73 deletions
--- a/backend/internal/service/gateway_forward_as_responses_test.go
+++ b/backend/internal/service/gateway_forward_as_responses_test.go
@@ -0,0 +1,94 @@
+//go:build unit
+
+package service
+
+import (
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/gin-gonic/gin"
+	"github.com/stretchr/testify/require"
+)
+
+func TestExtractResponsesReasoningEffortFromBody(t *testing.T) {
+	t.Parallel()
+
+	got := ExtractResponsesReasoningEffortFromBody([]byte(`{"model":"claude-sonnet-4.5","reasoning":{"effort":"HIGH"}}`))
+	require.NotNil(t, got)
+	require.Equal(t, "high", *got)
+
+	require.Nil(t, ExtractResponsesReasoningEffortFromBody([]byte(`{"model":"claude-sonnet-4.5"}`)))
+}
+
+func TestHandleResponsesBufferedStreamingResponse_PreservesMessageStartCacheUsage(t *testing.T) {
+	t.Parallel()
+	gin.SetMode(gin.TestMode)
+
+	rec := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(rec)
+
+	resp := &http.Response{
+		Header: http.Header{"x-request-id": []string{"rid_buffered"}},
+		Body: io.NopCloser(strings.NewReader(strings.Join([]string{
+			`event: message_start`,
+			`data: {"type":"message_start","message":{"id":"msg_1","type":"message","role":"assistant","content":[],"model":"claude-sonnet-4.5","stop_reason":"","usage":{"input_tokens":12,"cache_read_input_tokens":9,"cache_creation_input_tokens":3}}}`,
+			``,
+			`event: content_block_start`,
+			`data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":"hello"}}`,
+			``,
+			`event: message_delta`,
+			`data: {"type":"message_delta","delta":{"stop_reason":"end_turn"},"usage":{"output_tokens":7}}`,
+			``,
+		}, "\n"))),
+	}
+
+	svc := &GatewayService{}
+	result, err := svc.handleResponsesBufferedStreamingResponse(resp, c, "claude-sonnet-4.5", "claude-sonnet-4.5", nil, time.Now())
+	require.NoError(t, err)
+	require.NotNil(t, result)
+	require.Equal(t, 12, result.Usage.InputTokens)
+	require.Equal(t, 7, result.Usage.OutputTokens)
+	require.Equal(t, 9, result.Usage.CacheReadInputTokens)
+	require.Equal(t, 3, result.Usage.CacheCreationInputTokens)
+	require.Contains(t, rec.Body.String(), `"cached_tokens":9`)
+}
+
+func TestHandleResponsesStreamingResponse_PreservesMessageStartCacheUsage(t *testing.T) {
+	t.Parallel()
+	gin.SetMode(gin.TestMode)
+
+	rec := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(rec)
+
+	resp := &http.Response{
+		Header: http.Header{"x-request-id": []string{"rid_stream"}},
+		Body: io.NopCloser(strings.NewReader(strings.Join([]string{
+			`event: message_start`,
+			`data: {"type":"message_start","message":{"id":"msg_2","type":"message","role":"assistant","content":[],"model":"claude-sonnet-4.5","stop_reason":"","usage":{"input_tokens":20,"cache_read_input_tokens":11,"cache_creation_input_tokens":4}}}`,
+			``,
+			`event: content_block_start`,
+			`data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":"hello"}}`,
+			``,
+			`event: message_delta`,
+			`data: {"type":"message_delta","delta":{"stop_reason":"end_turn"},"usage":{"output_tokens":8}}`,
+			``,
+			`event: message_stop`,
+			`data: {"type":"message_stop"}`,
+			``,
+		}, "\n"))),
+	}
+
+	svc := &GatewayService{}
+	result, err := svc.handleResponsesStreamingResponse(resp, c, "claude-sonnet-4.5", "claude-sonnet-4.5", nil, time.Now())
+	require.NoError(t, err)
+	require.NotNil(t, result)
+	require.Equal(t, 20, result.Usage.InputTokens)
+	require.Equal(t, 8, result.Usage.OutputTokens)
+	require.Equal(t, 11, result.Usage.CacheReadInputTokens)
+	require.Equal(t, 4, result.Usage.CacheCreationInputTokens)
+	require.Contains(t, rec.Body.String(), `response.completed`)
+}