fix: Update API version and payload for dall-e-3 image generation

Author: Avijit-Microsoft

content-gen/src/backend/orchestrator.py

@@ -1245,7 +1245,14 @@ async def _generate_foundry_image(self, image_prompt: str, results: dict) -> Non
 
             # The direct image API endpoint
             image_api_url = f"{image_endpoint}/openai/deployments/{image_deployment}/images/generations"
-            api_version = app_settings.azure_openai.image_api_version or "2025-04-01-preview"
+
+            # Adapt API version and payload to the deployed image model
+            is_dalle3 = image_deployment.lower().startswith("dall-e")
+
+            if is_dalle3:
+                api_version = app_settings.azure_openai.preview_api_version or "2024-02-01"
+            else:
+                api_version = app_settings.azure_openai.image_api_version or "2025-04-01-preview"
 
             logger.info(f"Calling Foundry direct image API: {image_api_url}")
             logger.info(f"Prompt: {image_prompt[:200]}...")
@@ -1255,13 +1262,24 @@ async def _generate_foundry_image(self, image_prompt: str, results: dict) -> Non
                 "Content-Type": "application/json",
             }
 
-            # gpt-image-1 parameters (no response_format parameter)
-            payload = {
-                "prompt": image_prompt,
-                "n": 1,
-                "size": "1024x1024",
-                "quality": "medium",  # gpt-image-1 uses low/medium/high/auto
-            }
+            # Build model-appropriate payload
+            if is_dalle3:
+                # dall-e-3: quality must be "standard" or "hd"; needs response_format; 4000-char prompt limit
+                payload = {
+                    "prompt": image_prompt[:4000],
+                    "n": 1,
+                    "size": app_settings.azure_openai.image_size or "1024x1024",
+                    "quality": app_settings.azure_openai.image_quality or "hd",
+                    "response_format": "b64_json",
+                }
+            else:
+                # gpt-image-1 / gpt-image-1.5: quality is low/medium/high/auto; no response_format
+                payload = {
+                    "prompt": image_prompt,
+                    "n": 1,
+                    "size": "1024x1024",
+                    "quality": "medium",
+                }
 
             async with httpx.AsyncClient(timeout=120.0) as client:
                 response = await client.post(