tt/ttstd-infinite-canvas
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(*): 画布创作 - 支持文本多模态生成并优化媒体节点、连接线与视频预览

Browse Source 
Fixes #59
This commit is contained in:
ddCat
2026-07-31 19:44:02 +08:00
parent 5db39ccad4
commit 8de9b42c00
29 changed files with 231 additions and 111 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
backend/internal/service/provider_test.go
View File

@@ -117,13 +117,16 @@ func TestDoBinaryRejectsOversizedProviderResponse(t *testing.T) {
}
}
func TestTextResponseInputIncludesReferenceImages(t *testing.T) {
func TestTextResponseInputIncludesReferenceMedia(t *testing.T) {
input := canvasGenerationInput{
Prompt: "describe this image",
Config: providerConfig{SystemPrompt: "answer in Chinese"},
ReferenceImages: []providerMedia{
{ID: "image-1", Name: "image.png", Type: "image/png", DataURL: testReferenceImageDataURL},
},
ReferenceVideos: []providerMedia{
{ID: "video-1", Name: "video.mp4", Type: "video/mp4", URL: "https://example.com/reference.mp4"},
},
}
value, err := textResponseInput(input)
@@ -144,8 +147,8 @@ func TestTextResponseInputIncludesReferenceImages(t *testing.T) {
if !ok {
t.Fatalf("user content = %T, want []map[string]interface{}", messages[1]["content"])
}
if len(content) != 2 {
t.Fatalf("len(content) = %d, want 2", len(content))
if len(content) != 3 {
t.Fatalf("len(content) = %d, want 3", len(content))
}
if content[0]["type"] != "input_text" || content[0]["text"] != "describe this image" {
t.Fatalf("text content = %#v", content[0])
@@ -153,14 +156,20 @@ func TestTextResponseInputIncludesReferenceImages(t *testing.T) {
if content[1]["type"] != "input_image" || content[1]["image_url"] != testReferenceImageDataURL {
t.Fatalf("image content = %#v", content[1])
}
if content[2]["type"] != "input_video" || content[2]["video_url"] != "https://example.com/reference.mp4" {
t.Fatalf("video content = %#v", content[2])
}
}
func TestTextChatContentIncludesReferenceImages(t *testing.T) {
func TestTextChatContentIncludesReferenceMedia(t *testing.T) {
input := canvasGenerationInput{
Prompt: "describe this image",
ReferenceImages: []providerMedia{
{ID: "image-1", Name: "image.png", Type: "image/png", DataURL: testReferenceImageDataURL},
},
ReferenceVideos: []providerMedia{
{ID: "video-1", Name: "video.mp4", Type: "video/mp4", URL: "https://example.com/reference.mp4"},
},
}
value, err := textChatContent(input)
@@ -171,8 +180,8 @@ func TestTextChatContentIncludesReferenceImages(t *testing.T) {
if !ok {
t.Fatalf("textChatContent() = %T, want []map[string]interface{}", value)
}
if len(content) != 2 {
t.Fatalf("len(content) = %d, want 2", len(content))
if len(content) != 3 {
t.Fatalf("len(content) = %d, want 3", len(content))
}
if content[0]["type"] != "text" || content[0]["text"] != "describe this image" {
t.Fatalf("text content = %#v", content[0])
@@ -184,6 +193,13 @@ func TestTextChatContentIncludesReferenceImages(t *testing.T) {
if content[1]["type"] != "image_url" || imageURL["url"] != testReferenceImageDataURL {
t.Fatalf("image content = %#v", content[1])
}
videoURL, ok := content[2]["video_url"].(map[string]interface{})
if !ok {
t.Fatalf("video_url = %T, want map[string]interface{}", content[2]["video_url"])
}
if content[2]["type"] != "video_url" || videoURL["url"] != "https://example.com/reference.mp4" {
t.Fatalf("video content = %#v", content[2])
}
}
func TestTextReferenceImageRejectsInternalAssetURL(t *testing.T) {