fix(research): expose resolved model in API and logs (#4567)

routes/research/research_routes.py

@@ -147,6 +147,15 @@ def _resolve_research_endpoint(sess, owner: Optional[str] = None) -> tuple:
         fallback_headers=sess.headers,
         owner=owner,
     )
+    sess_model = getattr(sess, "model", None) or ""
+    if model and sess_model and model != sess_model:
+        logger.info(
+            "Research using configured model '%s' (session model was '%s')",
+            model,
+            sess_model,
+        )
+    elif model:
+        logger.info("Research using model '%s'", model)
     return url, model, headers
 
 
@@ -574,7 +583,7 @@ async def research_start(body: ResearchStartRequest, request: Request):
             extraction_concurrency=body.extraction_concurrency,
             owner=user,
         )
-        return {"session_id": session_id, "status": "running", "query": body.query}
+        return {"session_id": session_id, "status": "running", "query": body.query, "model": ep_model}
 
     @router.get("/api/research/stream/{session_id}")
     async def research_stream(session_id: str, request: Request):

src/research_handler.py

@@ -302,6 +302,7 @@ def start_research(
             "result": None,
             "started_at": time.time(),
             "category": category,
+            "llm_model": llm_model,
             # SECURITY: track ownership so all reads / saves can filter by user.
             "owner": owner or "",
         }
@@ -413,6 +414,7 @@ def get_status(self, session_id: str) -> Optional[dict]:
                 "progress": entry["progress"],
                 "query": entry["query"],
                 "started_at": entry["started_at"],
+                "model": entry.get("llm_model"),
             }
             # avg_duration is a historical figure over completed reports on
             # disk; get_avg_duration() globs and JSON-parses the whole research