fix(xai): rewrite image.image_url to image.url before upstream send

Normalize OpenAI-style image refs across images, videos, and Responses so Grok receives {image:{url:...}} instead of image_url.
2026-11-08 01:16:47 +08:00 · 2026-07-17 14:42:12 +08:00
parent 106270bea6
commit f82123a3ef
2 changed files with 189 additions and 5 deletions
@@ -497,13 +497,14 @@ func (e *XAIExecutor) executeImages(ctx context.Context, auth *cliproxyauth.Auth
 		endpointPath = xaiDefaultImageEndpointPath
 	}

+	payload := normalizeXAIImageRefs(req.Payload)
 	url := strings.TrimSuffix(baseURL, "/") + endpointPath
-	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(req.Payload))
+	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(payload))
 	if err != nil {
 		return resp, err
 	}
 	applyXAIHeaders(httpReq, auth, token, false, "")
-	e.recordXAIRequest(ctx, auth, url, httpReq.Header.Clone(), req.Payload)
+	e.recordXAIRequest(ctx, auth, url, httpReq.Header.Clone(), payload)

 	httpClient := helps.NewProxyAwareHTTPClient(ctx, e.cfg, auth, 0)
 	httpClient = reporter.TrackHTTPClient(httpClient)
@@ -543,15 +544,16 @@ func (e *XAIExecutor) executeVideos(ctx context.Context, auth *cliproxyauth.Auth
 	}
 	logXAIResolvedBaseURL(ctx, baseURL)

+	payload := normalizeXAIImageRefs(req.Payload)
 	method := http.MethodPost
 	endpointPath := xaiVideosGenerationsPath
-	var body io.Reader = bytes.NewReader(req.Payload)
+	var body io.Reader = bytes.NewReader(payload)

 	switch path := xaiVideoEndpointPath(opts); path {
 	case xaiVideosGenerationsPath, xaiVideosEditsPath, xaiVideosExtensionsPath:
 		endpointPath = path
 	default:
-		if requestID := strings.TrimSpace(gjson.GetBytes(req.Payload, "request_id").String()); requestID != "" {
+		if requestID := strings.TrimSpace(gjson.GetBytes(payload, "request_id").String()); requestID != "" {
 			method = http.MethodGet
 			endpointPath = xaiVideosPath + "/" + url.PathEscape(requestID)
 			body = nil
@@ -572,7 +574,7 @@ func (e *XAIExecutor) executeVideos(ctx context.Context, auth *cliproxyauth.Auth
 			httpReq.Header.Set("x-idempotency-key", key)
 		}
 	}
-	e.recordXAIRequest(ctx, auth, requestURL, httpReq.Header.Clone(), req.Payload)
+	e.recordXAIRequest(ctx, auth, requestURL, httpReq.Header.Clone(), payload)

 	httpClient := helps.NewProxyAwareHTTPClient(ctx, e.cfg, auth, 0)
 	httpResp, err := httpClient.Do(httpReq)
@@ -927,6 +929,7 @@ func (e *XAIExecutor) prepareResponsesRequestTo(ctx context.Context, req cliprox
 	body = sanitizeXAIInputEncryptedContent(body)
 	body = normalizeCodexInstructions(body)
 	body = sanitizeXAIResponsesBody(body, baseModel)
+	body = normalizeXAIImageRefs(body)

 	sessionID, errSession := xaiResolveComposerSessionID(ctx, req, opts, baseModel)
 	if errSession != nil {
@@ -1172,6 +1175,95 @@ func xaiImageEndpointPath(opts cliproxyexecutor.Options) string {
 	return xaiDefaultImageEndpointPath
 }

+// normalizeXAIImageRefs rewrites OpenAI-style image object fields to the xAI
+// image API shape before the payload is sent upstream:
+//
+//	{"image":{"image_url":"https://..."}} → {"image":{"url":"https://..."}}
+//
+// Applies to image / images / reference_images anywhere in the JSON tree,
+// including nested objects and array items. Does not rewrite chat content
+// parts shaped as {"type":"image_url","image_url":{...}}.
+func normalizeXAIImageRefs(body []byte) []byte {
+	if !gjson.ValidBytes(body) {
+		return body
+	}
+	return normalizeXAIImageRefsAt(body, "")
+}
+
+func normalizeXAIImageRefsAt(body []byte, path string) []byte {
+	node := gjson.GetBytes(body, path)
+	if path == "" {
+		node = gjson.ParseBytes(body)
+	}
+	if !node.Exists() {
+		return body
+	}
+
+	if node.IsObject() {
+		node.ForEach(func(key, value gjson.Result) bool {
+			childPath := key.String()
+			if path != "" {
+				childPath = path + "." + key.String()
+			}
+			switch key.String() {
+			case "image":
+				body = normalizeXAIImageRefAtPath(body, childPath)
+			case "images", "reference_images":
+				if value.IsArray() {
+					for i := range value.Array() {
+						body = normalizeXAIImageRefAtPath(body, fmt.Sprintf("%s.%d", childPath, i))
+					}
+				}
+			default:
+				if value.IsObject() || value.IsArray() {
+					body = normalizeXAIImageRefsAt(body, childPath)
+				}
+			}
+			return true
+		})
+		return body
+	}
+
+	if node.IsArray() {
+		for i := range node.Array() {
+			childPath := fmt.Sprintf("%d", i)
+			if path != "" {
+				childPath = fmt.Sprintf("%s.%d", path, i)
+			}
+			body = normalizeXAIImageRefsAt(body, childPath)
+		}
+	}
+	return body
+}
+
+func normalizeXAIImageRefAtPath(body []byte, path string) []byte {
+	ref := gjson.GetBytes(body, path)
+	if !ref.Exists() || ref.Type != gjson.JSON || !ref.IsObject() {
+		return body
+	}
+
+	url := strings.TrimSpace(ref.Get("url").String())
+	if url == "" {
+		if imageURL := ref.Get("image_url"); imageURL.Exists() {
+			if imageURL.Type == gjson.String {
+				url = strings.TrimSpace(imageURL.String())
+			} else if imageURL.IsObject() {
+				url = strings.TrimSpace(imageURL.Get("url").String())
+			}
+		}
+	}
+	if url == "" {
+		return body
+	}
+
+	// Always emit the xAI field name and drop the OpenAI alias.
+	body, _ = sjson.SetBytes(body, path+".url", url)
+	if ref.Get("image_url").Exists() {
+		body, _ = sjson.DeleteBytes(body, path+".image_url")
+	}
+	return body
+}
+
 func xaiIsVideoRequest(opts cliproxyexecutor.Options) bool {
 	return opts.SourceFormat.String() == xaiVideoHandlerType
 }
@@ -2250,6 +2250,98 @@ func TestXAIExecutorExecuteImagesUsesEditsEndpoint(t *testing.T) {
 	}
 }

+func TestNormalizeXAIImageRefsRewritesImageURLField(t *testing.T) {
+	t.Parallel()
+
+	in := []byte(`{
+		"model":"grok-imagine-image",
+		"prompt":"edit",
+		"image":{"type":"image_url","image_url":"https://example.com/a.png"},
+		"images":[{"image_url":{"url":"https://example.com/b.png"}},{"url":"https://example.com/c.png","image_url":"https://example.com/ignored.png"}],
+		"reference_images":[{"image_url":"https://example.com/d.png"}],
+		"nested":{"image":{"image_url":"https://example.com/e.png"}},
+		"content":[{"type":"image_url","image_url":{"url":"https://example.com/keep.png"}}]
+	}`)
+	out := normalizeXAIImageRefs(in)
+
+	if got := gjson.GetBytes(out, "image.url").String(); got != "https://example.com/a.png" {
+		t.Fatalf("image.url = %q, want https://example.com/a.png; body=%s", got, out)
+	}
+	if gjson.GetBytes(out, "image.image_url").Exists() {
+		t.Fatalf("image.image_url should be removed; body=%s", out)
+	}
+	if got := gjson.GetBytes(out, "image.type").String(); got != "image_url" {
+		t.Fatalf("image.type = %q, want image_url; body=%s", got, out)
+	}
+	if got := gjson.GetBytes(out, "images.0.url").String(); got != "https://example.com/b.png" {
+		t.Fatalf("images.0.url = %q, want https://example.com/b.png; body=%s", got, out)
+	}
+	if gjson.GetBytes(out, "images.0.image_url").Exists() {
+		t.Fatalf("images.0.image_url should be removed; body=%s", out)
+	}
+	if got := gjson.GetBytes(out, "images.1.url").String(); got != "https://example.com/c.png" {
+		t.Fatalf("images.1.url = %q, want existing url kept; body=%s", got, out)
+	}
+	if gjson.GetBytes(out, "images.1.image_url").Exists() {
+		t.Fatalf("images.1.image_url should be removed when url already set; body=%s", out)
+	}
+	if got := gjson.GetBytes(out, "reference_images.0.url").String(); got != "https://example.com/d.png" {
+		t.Fatalf("reference_images.0.url = %q, want https://example.com/d.png; body=%s", got, out)
+	}
+	if gjson.GetBytes(out, "reference_images.0.image_url").Exists() {
+		t.Fatalf("reference_images.0.image_url should be removed; body=%s", out)
+	}
+	if got := gjson.GetBytes(out, "nested.image.url").String(); got != "https://example.com/e.png" {
+		t.Fatalf("nested.image.url = %q, want https://example.com/e.png; body=%s", got, out)
+	}
+	if got := gjson.GetBytes(out, "content.0.image_url.url").String(); got != "https://example.com/keep.png" {
+		t.Fatalf("chat content image_url.url should be preserved, got %q; body=%s", got, out)
+	}
+	if gjson.GetBytes(out, "content.0.url").Exists() {
+		t.Fatalf("chat content parts must not be rewritten to url; body=%s", out)
+	}
+}
+
+func TestXAIExecutorExecuteImagesRewritesImageURLToURL(t *testing.T) {
+	var gotBody []byte
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		var errRead error
+		gotBody, errRead = io.ReadAll(r.Body)
+		if errRead != nil {
+			t.Fatalf("read body: %v", errRead)
+		}
+		w.Header().Set("Content-Type", "application/json")
+		_, _ = w.Write([]byte(`{"created":123,"data":[{"url":"https://x.ai/image.png"}]}`))
+	}))
+	defer server.Close()
+
+	exec := NewXAIExecutor(&config.Config{})
+	auth := &cliproxyauth.Auth{
+		Provider:   "xai",
+		Attributes: map[string]string{"base_url": server.URL},
+		Metadata:   map[string]any{"access_token": "xai-token"},
+	}
+
+	_, err := exec.Execute(context.Background(), auth, cliproxyexecutor.Request{
+		Model:   "grok-imagine-image",
+		Payload: []byte(`{"model":"grok-imagine-image","prompt":"edit","image":{"image_url":"https://example.com/a.png"}}`),
+	}, cliproxyexecutor.Options{
+		SourceFormat: sdktranslator.FromString("openai-image"),
+		Metadata: map[string]any{
+			cliproxyexecutor.RequestPathMetadataKey: "/v1/images/edits",
+		},
+	})
+	if err != nil {
+		t.Fatalf("Execute() error = %v", err)
+	}
+	if got := gjson.GetBytes(gotBody, "image.url").String(); got != "https://example.com/a.png" {
+		t.Fatalf("upstream image.url = %q, want https://example.com/a.png; body=%s", got, gotBody)
+	}
+	if gjson.GetBytes(gotBody, "image.image_url").Exists() {
+		t.Fatalf("upstream body still has image.image_url: %s", gotBody)
+	}
+}
+
 func TestXAIExecutorExecuteVideosCreate(t *testing.T) {
 	var gotPath string
 	var gotMethod string