feat(server): switch i2i to gpt (#12238)

fix AI-14
fix AI-17
fix AI-39
fix AI-112

<!-- This is an auto-generated comment: release notes by coderabbit.ai -->
## Summary by CodeRabbit

- **New Features**
  - Expanded and reorganized prompt options for text and image actions, adding new prompts for image generation, style conversions, upscaling, background removal, and sticker creation.
  - Enhanced image editing capabilities with direct support for image attachments in prompts.

- **Improvements**
  - Updated prompt names and descriptions to be more user-friendly and descriptive.
  - Simplified and clarified prompt selection and image processing workflows with improved default behaviors.
  - Better organization of prompts through clear grouping and categorization.

- **Bug Fixes**
  - Improved validation and handling of image attachments during editing requests.

- **Refactor**
  - Internal code restructuring of prompts and provider logic for clarity and maintainability without affecting user workflows.
  - Refined message handling and content merging logic to ensure consistent prompt processing.
  - Adjusted image attachment rendering logic for improved display consistency.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

packages/backend/server/src/__tests__/copilot-provider.spec.ts

@@ -518,12 +518,7 @@ const actions = [
     type: 'text' as const,
   },
   {
-    promptName: [
-      'debug:action:fal-face-to-sticker',
-      'debug:action:fal-remove-bg',
-      'debug:action:fal-sd15',
-      'debug:action:fal-upscaler',
-    ],
+    promptName: ['Convert to sticker', 'Remove background', 'Upscale image'],
     messages: [
       {
         role: 'user' as const,
@@ -590,6 +585,8 @@ for (const {
         }))!;
         t.truthy(provider, 'should have provider');
         await retry(`action: ${promptName}`, t, async t => {
+          const finalConfig = Object.assign({}, prompt.config, config);
+
           switch (type) {
             case 'text': {
               const result = await provider.text(
@@ -604,7 +601,7 @@ for (const {
                   ),
                   ...messages,
                 ],
-                Object.assign({}, prompt.config, config)
+                finalConfig
               );
               t.truthy(result, 'should return result');
               verifier?.(t, result);
@@ -622,23 +619,39 @@ for (const {
                   ),
                   ...messages,
                 ],
-                Object.assign({}, prompt.config, config)
+                finalConfig
               );
               t.truthy(result, 'should return result');
               verifier?.(t, result);
               break;
             }
             case 'image': {
-              const stream = provider.streamImages({ modelId: prompt.model }, [
-                ...prompt.finish(
-                  messages.reduce(
-                    // @ts-expect-error
-                    (acc, m) => Object.assign(acc, m.params),
-                    {}
-                  )
-                ),
-                ...messages,
-              ]);
+              const finalMessage = [...messages];
+              const params = {};
+              if (finalMessage.length === 1) {
+                const latestMessage = finalMessage.pop()!;
+                Object.assign(params, {
+                  content: latestMessage.content,
+                  attachments:
+                    'attachments' in latestMessage
+                      ? latestMessage.attachments
+                      : undefined,
+                });
+              }
+              const stream = provider.streamImages(
+                { modelId: prompt.model },
+                [
+                  ...prompt.finish(
+                    finalMessage.reduce(
+                      // @ts-expect-error
+                      (acc, m) => Object.assign(acc, m.params),
+                      params
+                    )
+                  ),
+                  ...finalMessage,
+                ],
+                finalConfig
+              );
 
               const result = [];
               for await (const attachment of stream) {

packages/backend/server/src/__tests__/copilot.e2e.ts

@@ -543,12 +543,19 @@ test('should be able to chat with special image model', async t => {
     );
   };
 
-  await testWithModel('debug:action:fal-sd15', 'some-tag');
+  await testWithModel('Generate image', 'some-tag');
   await testWithModel(
-    'debug:action:fal-upscaler',
-    'best quality, 8K resolution, highres, clarity, some-tag'
+    'Convert to sticker',
+    'convert this image to sticker. you need to identify the subject matter and warp a circle of white stroke around the subject matter and with transparent background. some-tag'
+  );
+  await testWithModel(
+    'Upscale image',
+    'make the image more detailed. some-tag'
+  );
+  await testWithModel(
+    'Remove background',
+    'Keep the subject and remove other non-subject items. Transparent background. some-tag'
   );
-  await testWithModel('debug:action:fal-remove-bg', 'some-tag');
 
   Sinon.restore();
 });

packages/backend/server/src/__tests__/mocks/copilot.mock.ts

@@ -84,29 +84,12 @@ export class MockCopilotProvider extends OpenAIProvider {
       ],
     },
     {
-      id: 'lcm-sd15-i2i',
+      id: 'gpt-image-1',
       capabilities: [
         {
-          input: [ModelInputType.Image],
-          output: [ModelOutputType.Image],
-        },
-      ],
-    },
-    {
-      id: 'clarity-upscaler',
-      capabilities: [
-        {
-          input: [ModelInputType.Image],
-          output: [ModelOutputType.Image],
-        },
-      ],
-    },
-    {
-      id: 'imageutils/rembg',
-      capabilities: [
-        {
-          input: [ModelInputType.Image],
+          input: [ModelInputType.Text, ModelInputType.Image],
           output: [ModelOutputType.Image],
+          defaultForOutputType: true,
         },
       ],
     },