Add experimental image generation fields to /api/generate (#13753)
Request fields (experimental): - width: image width (max 4096) - height: image height (max 4096) - steps: denoising steps - seed: random seed Response fields (experimental): - images: base64-encoded generated images - completed: current step progress - total: total steps Other changes: - Fix lifecycle bug where image models wouldn't unload (refCount issue) - Fix "headers already written" error on Ctrl+C during streaming - Add gin middleware for OpenAI /v1/images/generations compatibility - Update CLI to use /api/generate with progress bar - Add preload support in interactive mode
This commit is contained in:
Jeffrey Morgan
@@ -546,3 +546,66 @@ func ResponsesMiddleware() gin.HandlerFunc {
|
||||
c.Next()
|
||||
}
|
||||
}
|
||||
|
||||
type ImageWriter struct {
|
||||
BaseWriter
|
||||
}
|
||||
|
||||
func (w *ImageWriter) writeResponse(data []byte) (int, error) {
|
||||
var generateResponse api.GenerateResponse
|
||||
if err := json.Unmarshal(data, &generateResponse); err != nil {
|
||||
return 0, err
|
||||
}
|
||||
|
||||
// Only write response when done with image
|
||||
if generateResponse.Done && generateResponse.Image != "" {
|
||||
w.ResponseWriter.Header().Set("Content-Type", "application/json")
|
||||
return len(data), json.NewEncoder(w.ResponseWriter).Encode(openai.ToImageGenerationResponse(generateResponse))
|
||||
}
|
||||
|
||||
return len(data), nil
|
||||
}
|
||||
|
||||
func (w *ImageWriter) Write(data []byte) (int, error) {
|
||||
code := w.ResponseWriter.Status()
|
||||
if code != http.StatusOK {
|
||||
return w.writeError(data)
|
||||
}
|
||||
|
||||
return w.writeResponse(data)
|
||||
}
|
||||
|
||||
func ImageGenerationsMiddleware() gin.HandlerFunc {
|
||||
return func(c *gin.Context) {
|
||||
var req openai.ImageGenerationRequest
|
||||
if err := c.ShouldBindJSON(&req); err != nil {
|
||||
c.AbortWithStatusJSON(http.StatusBadRequest, openai.NewError(http.StatusBadRequest, err.Error()))
|
||||
return
|
||||
}
|
||||
|
||||
if req.Prompt == "" {
|
||||
c.AbortWithStatusJSON(http.StatusBadRequest, openai.NewError(http.StatusBadRequest, "prompt is required"))
|
||||
return
|
||||
}
|
||||
|
||||
if req.Model == "" {
|
||||
c.AbortWithStatusJSON(http.StatusBadRequest, openai.NewError(http.StatusBadRequest, "model is required"))
|
||||
return
|
||||
}
|
||||
|
||||
var b bytes.Buffer
|
||||
if err := json.NewEncoder(&b).Encode(openai.FromImageGenerationRequest(req)); err != nil {
|
||||
c.AbortWithStatusJSON(http.StatusInternalServerError, openai.NewError(http.StatusInternalServerError, err.Error()))
|
||||
return
|
||||
}
|
||||
|
||||
c.Request.Body = io.NopCloser(&b)
|
||||
|
||||
w := &ImageWriter{
|
||||
BaseWriter: BaseWriter{ResponseWriter: c.Writer},
|
||||
}
|
||||
|
||||
c.Writer = w
|
||||
c.Next()
|
||||
}
|
||||
}
|
||||
|
||||
@@ -961,3 +961,154 @@ func TestRetrieveMiddleware(t *testing.T) {
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func TestImageGenerationsMiddleware(t *testing.T) {
|
||||
type testCase struct {
|
||||
name string
|
||||
body string
|
||||
req api.GenerateRequest
|
||||
err openai.ErrorResponse
|
||||
}
|
||||
|
||||
var capturedRequest *api.GenerateRequest
|
||||
|
||||
testCases := []testCase{
|
||||
{
|
||||
name: "image generation basic",
|
||||
body: `{
|
||||
"model": "test-model",
|
||||
"prompt": "a beautiful sunset"
|
||||
}`,
|
||||
req: api.GenerateRequest{
|
||||
Model: "test-model",
|
||||
Prompt: "a beautiful sunset",
|
||||
},
|
||||
},
|
||||
{
|
||||
name: "image generation with size",
|
||||
body: `{
|
||||
"model": "test-model",
|
||||
"prompt": "a beautiful sunset",
|
||||
"size": "512x768"
|
||||
}`,
|
||||
req: api.GenerateRequest{
|
||||
Model: "test-model",
|
||||
Prompt: "a beautiful sunset",
|
||||
Width: 512,
|
||||
Height: 768,
|
||||
},
|
||||
},
|
||||
{
|
||||
name: "image generation missing prompt",
|
||||
body: `{
|
||||
"model": "test-model"
|
||||
}`,
|
||||
err: openai.ErrorResponse{
|
||||
Error: openai.Error{
|
||||
Message: "prompt is required",
|
||||
Type: "invalid_request_error",
|
||||
},
|
||||
},
|
||||
},
|
||||
{
|
||||
name: "image generation missing model",
|
||||
body: `{
|
||||
"prompt": "a beautiful sunset"
|
||||
}`,
|
||||
err: openai.ErrorResponse{
|
||||
Error: openai.Error{
|
||||
Message: "model is required",
|
||||
Type: "invalid_request_error",
|
||||
},
|
||||
},
|
||||
},
|
||||
}
|
||||
|
||||
endpoint := func(c *gin.Context) {
|
||||
c.Status(http.StatusOK)
|
||||
}
|
||||
|
||||
gin.SetMode(gin.TestMode)
|
||||
router := gin.New()
|
||||
router.Use(ImageGenerationsMiddleware(), captureRequestMiddleware(&capturedRequest))
|
||||
router.Handle(http.MethodPost, "/api/generate", endpoint)
|
||||
|
||||
for _, tc := range testCases {
|
||||
t.Run(tc.name, func(t *testing.T) {
|
||||
req, _ := http.NewRequest(http.MethodPost, "/api/generate", strings.NewReader(tc.body))
|
||||
req.Header.Set("Content-Type", "application/json")
|
||||
|
||||
defer func() { capturedRequest = nil }()
|
||||
|
||||
resp := httptest.NewRecorder()
|
||||
router.ServeHTTP(resp, req)
|
||||
|
||||
if tc.err.Error.Message != "" {
|
||||
var errResp openai.ErrorResponse
|
||||
if err := json.Unmarshal(resp.Body.Bytes(), &errResp); err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
if diff := cmp.Diff(tc.err, errResp); diff != "" {
|
||||
t.Fatalf("errors did not match:\n%s", diff)
|
||||
}
|
||||
return
|
||||
}
|
||||
|
||||
if resp.Code != http.StatusOK {
|
||||
t.Fatalf("expected status 200, got %d: %s", resp.Code, resp.Body.String())
|
||||
}
|
||||
|
||||
if diff := cmp.Diff(&tc.req, capturedRequest); diff != "" {
|
||||
t.Fatalf("requests did not match:\n%s", diff)
|
||||
}
|
||||
})
|
||||
}
|
||||
}
|
||||
|
||||
func TestImageWriterResponse(t *testing.T) {
|
||||
gin.SetMode(gin.TestMode)
|
||||
|
||||
// Test that ImageWriter transforms GenerateResponse to OpenAI format
|
||||
endpoint := func(c *gin.Context) {
|
||||
resp := api.GenerateResponse{
|
||||
Model: "test-model",
|
||||
CreatedAt: time.Unix(1234567890, 0).UTC(),
|
||||
Done: true,
|
||||
Image: "dGVzdC1pbWFnZS1kYXRh", // base64 of "test-image-data"
|
||||
}
|
||||
data, _ := json.Marshal(resp)
|
||||
c.Writer.Write(append(data, '\n'))
|
||||
}
|
||||
|
||||
router := gin.New()
|
||||
router.Use(ImageGenerationsMiddleware())
|
||||
router.Handle(http.MethodPost, "/api/generate", endpoint)
|
||||
|
||||
body := `{"model": "test-model", "prompt": "test"}`
|
||||
req, _ := http.NewRequest(http.MethodPost, "/api/generate", strings.NewReader(body))
|
||||
req.Header.Set("Content-Type", "application/json")
|
||||
|
||||
resp := httptest.NewRecorder()
|
||||
router.ServeHTTP(resp, req)
|
||||
|
||||
if resp.Code != http.StatusOK {
|
||||
t.Fatalf("expected status 200, got %d: %s", resp.Code, resp.Body.String())
|
||||
}
|
||||
|
||||
var imageResp openai.ImageGenerationResponse
|
||||
if err := json.Unmarshal(resp.Body.Bytes(), &imageResp); err != nil {
|
||||
t.Fatalf("failed to unmarshal response: %v", err)
|
||||
}
|
||||
|
||||
if imageResp.Created != 1234567890 {
|
||||
t.Errorf("expected created 1234567890, got %d", imageResp.Created)
|
||||
}
|
||||
|
||||
if len(imageResp.Data) != 1 {
|
||||
t.Fatalf("expected 1 image, got %d", len(imageResp.Data))
|
||||
}
|
||||
|
||||
if imageResp.Data[0].B64JSON != "dGVzdC1pbWFnZS1kYXRh" {
|
||||
t.Errorf("expected image data 'dGVzdC1pbWFnZS1kYXRh', got %s", imageResp.Data[0].B64JSON)
|
||||
}
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user