feat(phase4): permanent cache-control & asset versioning, canonical nabil stored_session_id & resume, dsh session persistence fix, dsh dialogue/files toggle, and collapsible conv sidebar

app/chat_service.py

@@ -114,7 +114,7 @@ async def list_universe_conversations(universe_id: str) -> List[Dict[str, Any]]:
                     for s in backend_sessions:
                         sid = str(s.get("id") or s.get("session_id") or "")
                         if sid:
-                            stitle = s.get("title") or s.get("display_name") or f"Session {sid[:8]}"
+                            stitle = s.get("title") or s.get("display_name") or f"Session {sid[:15]}"
                             spinned = 1 if s.get("pinned") else 0
                             started_at = s.get("started_at")
                             c_at = int(started_at * 1000) if isinstance(started_at, (int, float)) and started_at < 1e11 else int(time.time() * 1000)
@@ -171,9 +171,10 @@ async def create_universe_conversation(universe_id: str, title: Optional[str] =
                     data = json.loads(raw)
                     if data.get("id") == req_id and "result" in data:
                         res = data["result"]
-                        session_key = res.get("session_id") or res.get("stored_session_id")
+                        # Prioritize durable stored_session_id (format YYYYMMDD_HHMMSS_xxxxxx) over ephemeral handle
+                        session_key = res.get("stored_session_id") or res.get("session_id")
                         break
-        except Exception as e:
+        except Exception:
             # Fallback if WS fails
             session_key = f"nabil_{uuid.uuid4().hex[:12]}"
             
@@ -234,7 +235,6 @@ async def get_conversation_messages(universe_id: str, session_key: str) -> List[
 
 async def stream_chat_messages(universe_id: str, session_key: str, message: str) -> AsyncGenerator[str, None]:
     """Streams chat chunks using Server-Sent Events (SSE)."""
-    # Touch conversation in SQLite and update title if first message
     existing = get_conversation(universe_id, session_key)
     if existing and existing.get("title") in ("Nouvelle conversation", "Untitled", None):
         summary_title = (message[:38] + "...") if len(message) > 38 else message
@@ -272,13 +272,35 @@ async def stream_chat_messages(universe_id: str, session_key: str, message: str)
                 ping_interval=20,
                 ping_timeout=20
             ) as ws:
+                # 1. First resume/activate the session to get the active live handle
+                resume_req_id = f"res-{uuid.uuid4().hex[:8]}"
+                await ws.send(json.dumps({
+                    "jsonrpc": "2.0",
+                    "id": resume_req_id,
+                    "method": "session.resume",
+                    "params": {"session_id": session_key}
+                }))
+                
+                active_sid = session_key
+                try:
+                    while True:
+                        raw_res = await asyncio.wait_for(ws.recv(), timeout=4.0)
+                        data_res = json.loads(raw_res)
+                        if data_res.get("id") == resume_req_id:
+                            if "result" in data_res:
+                                active_sid = data_res["result"].get("session_id") or session_key
+                            break
+                except Exception:
+                    pass
+                
+                # 2. Submit prompt with active_sid
                 req_id = f"prompt-{uuid.uuid4().hex[:8]}"
                 send_payload = {
                     "jsonrpc": "2.0",
                     "id": req_id,
                     "method": "prompt.submit",
                     "params": {
-                        "session_id": session_key,
+                        "session_id": active_sid,
                         "text": message
                     }
                 }
@@ -300,7 +322,6 @@ async def stream_chat_messages(universe_id: str, session_key: str, message: str)
                             yield f"event: chunk\ndata: {json.dumps({'text': accumulated_text, 'chunk': delta, 'fullReplace': True})}\n\n"
                         elif ev_type in ("reasoning.delta", "thinking.delta"):
                             delta = payload.get("text", "")
-                            # We can stream reasoning as needed
                         elif ev_type in ("message.finish", "session.idle", "reasoning.available") or data.get("method") in ("chat.done", "session.done"):
                             yield f"event: done\ndata: {json.dumps({'state': 'complete', 'text': accumulated_text})}\n\n"
                             break