feat: add composite message index and reduce pool timeout (#5156)

alembic/versions/f9ad1c25fd2b_add_query_optimizing_runs_listing.py

@@ -0,0 +1,31 @@
+"""add query optimizing runs listing
+
+Revision ID: f9ad1c25fd2b
+Revises: 3bc3c031fbe4
+Create Date: 2025-10-04 00:44:06.663817
+
+"""
+
+from typing import Sequence, Union
+
+import sqlalchemy as sa
+
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision: str = "f9ad1c25fd2b"
+down_revision: Union[str, None] = "3bc3c031fbe4"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.create_index("ix_messages_run_err_sequence", "messages", ["run_id", "is_err", "sequence_id"], unique=False)
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_index("ix_messages_run_err_sequence", table_name="messages")
+    # ### end Alembic commands ###

letta/orm/message.py

@@ -22,6 +22,8 @@ class Message(SqlalchemyBase, OrganizationMixin, AgentMixin):
         Index("ix_messages_agent_sequence", "agent_id", "sequence_id"),
         Index("ix_messages_org_agent", "organization_id", "agent_id"),
         Index("ix_messages_run_id", "run_id"),
+        # Composite index for optimizing the frequently-run query:
+        Index("ix_messages_run_err_sequence", "run_id", "is_err", "sequence_id"),
     )
     __pydantic_model__ = PydanticMessage
 

letta/settings.py

@@ -236,7 +236,7 @@ class Settings(BaseSettings):
     pg_uri: Optional[str] = default_pg_uri  # option to specify full uri
     pg_pool_size: int = 25  # Concurrent connections
     pg_max_overflow: int = 10  # Overflow limit
-    pg_pool_timeout: int = 90  # Seconds to wait for a connection
+    pg_pool_timeout: int = 30  # Seconds to wait for a connection
     pg_pool_recycle: int = 1800  # When to recycle connections
     pg_echo: bool = False  # Logging
     pool_pre_ping: bool = True  # Pre ping to check for dead connections