fix: properly limit runs query (#6357)

* letta

* tweak

* cancellation to 100

* openapi changes

fern/openapi.json

@@ -10782,7 +10782,8 @@
               "anyOf": [
                 {
                   "type": "integer",
-                  "maximum": 1000
+                  "maximum": 1000,
+                  "minimum": 1
                 },
                 {
                   "type": "null"
@@ -13722,7 +13723,8 @@
               "anyOf": [
                 {
                   "type": "integer",
-                  "maximum": 1000
+                  "maximum": 1000,
+                  "minimum": 1
                 },
                 {
                   "type": "null"

letta/server/rest_api/routers/v1/agents.py

@@ -1596,6 +1596,7 @@ async def cancel_message(
                 statuses=[RunStatus.created, RunStatus.running],
                 ascending=False,
                 agent_id=agent_id,  # NOTE: this will override agent_ids if provided
+                limit=100,  # Limit to 10 most recent active runs for cancellation
             )
             run_ids = [run.id for run in run_ids]
         else:

letta/server/rest_api/routers/v1/internal_runs.py

@@ -52,7 +52,7 @@ async def list_runs(
     after: Optional[str] = Query(
         None, description="Run ID cursor for pagination. Returns runs that come after this run ID in the specified sort order"
     ),
-    limit: Optional[int] = Query(100, description="Maximum number of runs to return", le=1000),
+    limit: Optional[int] = Query(100, description="Maximum number of runs to return", ge=1, le=1000),
     order: Literal["asc", "desc"] = Query(
         "desc", description="Sort order for runs by creation time. 'asc' for oldest first, 'desc' for newest first"
     ),

letta/server/rest_api/routers/v1/runs.py

@@ -61,7 +61,7 @@ async def list_runs(
     after: Optional[str] = Query(
         None, description="Run ID cursor for pagination. Returns runs that come after this run ID in the specified sort order"
     ),
-    limit: Optional[int] = Query(100, description="Maximum number of runs to return", le=1000),
+    limit: Optional[int] = Query(100, description="Maximum number of runs to return", ge=1, le=1000),
     order: Literal["asc", "desc"] = Query(
         "desc", description="Sort order for runs by creation time. 'asc' for oldest first, 'desc' for newest first"
     ),
@@ -132,7 +132,7 @@ async def list_active_runs(
         agent_ids = None
 
     active_runs = await runs_manager.list_runs(
-        actor=actor, statuses=[RunStatus.created, RunStatus.running], agent_ids=agent_ids, background=background
+        actor=actor, statuses=[RunStatus.created, RunStatus.running], agent_ids=agent_ids, background=background, limit=100
     )
 
     return active_runs

letta/services/run_manager.py

@@ -267,9 +267,11 @@ class RunManager:
 
                 query = await _apply_pagination_async(query, before, after, session, ascending=ascending)
 
-            # Apply limit
-            if limit:
-                query = query.limit(limit)
+            # Apply limit (always enforce a maximum to prevent unbounded queries)
+            # If no limit specified, default to 100; enforce maximum of 1000
+            effective_limit = limit if limit is not None else 100
+            effective_limit = min(effective_limit, 1000)
+            query = query.limit(effective_limit)
 
             result = await session.execute(query)
             rows = result.all()