mirror of
https://github.com/ollama/ollama.git
synced 2026-07-23 09:10:53 -05:00
x/imagegen: add image edit capabilities (#13846)
This commit is contained in:
@@ -794,3 +794,47 @@ func ToImageGenerationResponse(resp api.GenerateResponse) ImageGenerationRespons
|
||||
Data: data,
|
||||
}
|
||||
}
|
||||
|
||||
// ImageEditRequest is an OpenAI-compatible image edit request.
|
||||
type ImageEditRequest struct {
|
||||
Model string `json:"model"`
|
||||
Prompt string `json:"prompt"`
|
||||
Image string `json:"image"` // Base64-encoded image data
|
||||
Size string `json:"size,omitempty"` // e.g., "1024x1024"
|
||||
Seed *int64 `json:"seed,omitempty"`
|
||||
}
|
||||
|
||||
// FromImageEditRequest converts an OpenAI image edit request to an Ollama GenerateRequest.
|
||||
func FromImageEditRequest(r ImageEditRequest) (api.GenerateRequest, error) {
|
||||
req := api.GenerateRequest{
|
||||
Model: r.Model,
|
||||
Prompt: r.Prompt,
|
||||
}
|
||||
|
||||
// Decode the input image
|
||||
if r.Image != "" {
|
||||
imgData, err := decodeImageURL(r.Image)
|
||||
if err != nil {
|
||||
return api.GenerateRequest{}, fmt.Errorf("invalid image: %w", err)
|
||||
}
|
||||
req.Images = append(req.Images, imgData)
|
||||
}
|
||||
|
||||
// Parse size if provided (e.g., "1024x768")
|
||||
if r.Size != "" {
|
||||
var w, h int32
|
||||
if _, err := fmt.Sscanf(r.Size, "%dx%d", &w, &h); err == nil {
|
||||
req.Width = w
|
||||
req.Height = h
|
||||
}
|
||||
}
|
||||
|
||||
if r.Seed != nil {
|
||||
if req.Options == nil {
|
||||
req.Options = map[string]any{}
|
||||
}
|
||||
req.Options["seed"] = *r.Seed
|
||||
}
|
||||
|
||||
return req, nil
|
||||
}
|
||||
|
||||
@@ -448,3 +448,86 @@ func TestFromChatRequest_TopLogprobsRange(t *testing.T) {
|
||||
})
|
||||
}
|
||||
}
|
||||
|
||||
func TestFromImageEditRequest_Basic(t *testing.T) {
|
||||
req := ImageEditRequest{
|
||||
Model: "test-model",
|
||||
Prompt: "make it blue",
|
||||
Image: prefix + image,
|
||||
}
|
||||
|
||||
result, err := FromImageEditRequest(req)
|
||||
if err != nil {
|
||||
t.Fatalf("unexpected error: %v", err)
|
||||
}
|
||||
|
||||
if result.Model != "test-model" {
|
||||
t.Errorf("expected model 'test-model', got %q", result.Model)
|
||||
}
|
||||
|
||||
if result.Prompt != "make it blue" {
|
||||
t.Errorf("expected prompt 'make it blue', got %q", result.Prompt)
|
||||
}
|
||||
|
||||
if len(result.Images) != 1 {
|
||||
t.Fatalf("expected 1 image, got %d", len(result.Images))
|
||||
}
|
||||
}
|
||||
|
||||
func TestFromImageEditRequest_WithSize(t *testing.T) {
|
||||
req := ImageEditRequest{
|
||||
Model: "test-model",
|
||||
Prompt: "make it blue",
|
||||
Image: prefix + image,
|
||||
Size: "512x768",
|
||||
}
|
||||
|
||||
result, err := FromImageEditRequest(req)
|
||||
if err != nil {
|
||||
t.Fatalf("unexpected error: %v", err)
|
||||
}
|
||||
|
||||
if result.Width != 512 {
|
||||
t.Errorf("expected width 512, got %d", result.Width)
|
||||
}
|
||||
|
||||
if result.Height != 768 {
|
||||
t.Errorf("expected height 768, got %d", result.Height)
|
||||
}
|
||||
}
|
||||
|
||||
func TestFromImageEditRequest_WithSeed(t *testing.T) {
|
||||
seed := int64(12345)
|
||||
req := ImageEditRequest{
|
||||
Model: "test-model",
|
||||
Prompt: "make it blue",
|
||||
Image: prefix + image,
|
||||
Seed: &seed,
|
||||
}
|
||||
|
||||
result, err := FromImageEditRequest(req)
|
||||
if err != nil {
|
||||
t.Fatalf("unexpected error: %v", err)
|
||||
}
|
||||
|
||||
if result.Options == nil {
|
||||
t.Fatal("expected options to be set")
|
||||
}
|
||||
|
||||
if result.Options["seed"] != seed {
|
||||
t.Errorf("expected seed %d, got %v", seed, result.Options["seed"])
|
||||
}
|
||||
}
|
||||
|
||||
func TestFromImageEditRequest_InvalidImage(t *testing.T) {
|
||||
req := ImageEditRequest{
|
||||
Model: "test-model",
|
||||
Prompt: "make it blue",
|
||||
Image: "not-valid-base64",
|
||||
}
|
||||
|
||||
_, err := FromImageEditRequest(req)
|
||||
if err == nil {
|
||||
t.Error("expected error for invalid image")
|
||||
}
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user