Open-source AI teammates in Slack, Teams, Discord, Telegram and Mattermost. Self-host it, with any agent harness and any model.
Python agent-orchestrationaiai-agentschatopsdiscordlangchainmcpmicrosoft-teamsmulti-agentopen-sourceopenaiself-hostedslack
Index room_agents on agent_id (CHOO-3667)
Abel Dantas6c7ccab9a9e6
diff --git a/CHANGELOG.md b/CHANGELOG.mdindex 8b9791612..4779b0e5f 100644--- a/CHANGELOG.md+++ b/CHANGELOG.md@@ -224,6 +224,9 @@ version of their own to them without also giving them a release of their own.group whose subgroup would move up next to a namesake is refused too, andthe dashboard now shows why. Names can no longer contain `/` or be blank.Existing duplicates are left as they are and can be reached by id.+- **Reading an agent's rooms uses an index.** `room_agents` is indexed on+ `agent_id`; reads by agent scanned the whole table, 2.9 ms each at 10,000+ agents.#### Fixed- **On Slack, the 👀 and the activity card no longer arrive after the agent'sdiff --git a/core/switch_core/db/models.py b/core/switch_core/db/models.pyindex 70dff1e2c..807b1e746 100644--- a/core/switch_core/db/models.py+++ b/core/switch_core/db/models.py@@ -886,6 +886,9 @@ room_agents = Table(name="fk_room_agents_agent",),Index("ix_room_agents_tenant_id", "tenant_id"),+ # The primary key leads with the room; reads of an agent's rooms, and the+ # foreign key's check when an agent is deleted, go by the agent.+ Index("ix_room_agents_agent_id", "agent_id"),)room_skills = Table(diff --git a/core/switch_core/migrations/versions/063652bcb450_index_room_agents_agent_id.py b/core/switch_core/migrations/versions/063652bcb450_index_room_agents_agent_id.pynew file mode 100644index 000000000..14d4c3a61--- /dev/null+++ b/core/switch_core/migrations/versions/063652bcb450_index_room_agents_agent_id.py@@ -0,0 +1,44 @@+"""index room_agents on agent_id++Reading the rooms an agent is in (`RoomStore.get_rooms_for_agent` and six+other reads by agent), and the check of `fk_room_agents_agent` when an agent+is deleted, filter `room_agents` by agent. The primary key is (room_id,+agent_id), whose leading column is the room, so none could use it and each+was a sequential scan: 2.9 ms at 10,000 agents, and reading the rooms of a+controller's agents, which every process does when a controller connects,+ran it for each of them.++Not `CONCURRENTLY`, for the reason `b8f2d0c41e57` gives: revisions run in+a transaction, and the table is small on every deployment today. A deployment+where it is not can build the index by hand first, and this becomes a no-op:++ CREATE INDEX CONCURRENTLY ix_room_agents_agent_id ON room_agents (agent_id);++Expand only: a process from before this revision reads the table as it did.++Revision ID: 063652bcb450+Revises: b7e1d4c9a2f6+Create Date: 2026-11-01 00:00:00.000000++"""++from collections.abc import Sequence++from alembic import op++revision: str = "063652bcb450"+down_revision: str | None = "b7e1d4c9a2f6"+branch_labels: str | Sequence[str] | None = None+depends_on: str | Sequence[str] | None = None+expand_only: bool = True++_INDEX = "ix_room_agents_agent_id"+++def upgrade() -> None:+ # `IF NOT EXISTS` so an index built by hand ahead of the upgrade is kept.+ op.execute(f"CREATE INDEX IF NOT EXISTS {_INDEX} ON room_agents (agent_id)")+++def downgrade() -> None:+ op.execute(f"DROP INDEX IF EXISTS {_INDEX}")