Merge feat/multi-agent into development

- Resolved conflict in api/app/agents/memoir/prompts.py: kept archived_section for get_narrative_prompt compatibility
- Resolved conflict in api/app/tasks/memoir_tasks.py: use MemoirOrchestrator flow (feat/multi-agent)
- Multi-agent architecture: ChatOrchestrator, MemoirOrchestrator, ImagePromptOrchestrator

api/app/features/memoir/memoir_images/prompting.py

@@ -57,7 +57,11 @@ class MemoirImagePromptService:
         if self.llm:
             raw_response = None
             try:
-                response = self.llm.invoke(
+                json_llm = self.llm.bind(
+                    model_kwargs={"response_format": {"type": "json_object"}},
+                    max_tokens=512,
+                )
+                response = json_llm.invoke(
                     "Return JSON only with keys prompt, style, size. "
                     "Convert the memoir scene into an image-generation prompt.\n"
                     + json.dumps(llm_input, ensure_ascii=False)
@@ -113,7 +117,11 @@ class MemoirImagePromptService:
 
         if self.llm:
             try:
-                response = self.llm.invoke(
+                json_llm = self.llm.bind(
+                    model_kwargs={"response_format": {"type": "json_object"}},
+                    max_tokens=512,
+                )
+                response = json_llm.invoke(
                     "Return JSON only with keys prompt, style, size. "
                     "Create an image-generation prompt for a memoir chapter COVER. "
                     "Emphasize: hero composition, evocative scene, chapter cover aesthetic.\n"