Response Lag
Hey team — question about reducing agent response latency.Current setup: Agent picks up messages from a Supabase table via a cron running every 60s.
Median response time is ~71s. Most of that is just waiting for the next cron tick — LLM processing
is only 3-10s.
What I want: Near-instant pickup. Message lands in DB, agent sees it within 1-2s.
Options I'm considering:
Reduce cron interval to 15s (easy but still polling)
Agent subscribes to Supabase Realtime for INSERT events — responds immediately
Database webhook triggers the agent on INSERT instead of waiting for cron
Questions:
Does OpenClaw support long-running Supabase Realtime subscriptions on the VPS? Or would a
persistent websocket get killed?
Is there a way to trigger a skill/cron immediately on an external event (webhook) instead of
waiting for the next tick?
What's the recommended pattern for low-latency message handling?
Agent writes responses back into Supabase (same team_messages table, channel='direct').
Plan: run a small Realtime subscriber relay on the VPS that calls /hooks/agent on INSERT. Keep
cron polling as dead-letter fallback at reduced frequency.
Two questions:
For the /hooks/agent payload, should the "message" field contain the raw user message content,
or should I include metadata (from_agent_id, message ID) so the agent can query the full context?
With sessionKey per conversation thread — does OpenClaw coalesce multiple webhook calls into one
agent turn, or does each call start a separate run?
The AI that actually does things. Emails, calendar, home automation - all from your favorite chat app. New shell, ready to help. The lobster way.
83,815Members
Resources
Similar Threads
Was this page helpful?
