Problem
claw-wall has a default Discord poll interval of 30 seconds (defaultWallPollSeconds = 30 in cmd/claw-wall/main.go). This is already configurable via CLAW_WALL_POLL_INTERVAL, but the default is higher than it should be.
The intended update frequency for channel context is ~15 seconds. At 30s, agents can be working from context that's up to half a minute stale.
Background
claw-wall uses a background poller pattern — all agents share one in-memory conversationStore, and /channel-context requests never hit Discord directly. So poll interval is the only knob that controls freshness.
The default of 30s was probably conservative, but in practice the trading desk wants 15s channel freshness. The CLAW_WALL_POLL_INTERVAL env var already exists as the escape hatch, but the default should reflect the intended cadence.
Proposed Change
- Change
defaultWallPollSeconds from 30 to 15 in cmd/claw-wall/main.go
- Update any docs/tests that reference the 30s default
Notes
While investigating a sentinel alert ("weston feed error rate 50.0%"), confirmed that the connection resets in claw-wall logs are transient Discord TCP failures — not caused by poll frequency. Stale Discord context is a separate, lower-severity issue this change would also improve.
Problem
claw-wallhas a default Discord poll interval of 30 seconds (defaultWallPollSeconds = 30incmd/claw-wall/main.go). This is already configurable viaCLAW_WALL_POLL_INTERVAL, but the default is higher than it should be.The intended update frequency for channel context is ~15 seconds. At 30s, agents can be working from context that's up to half a minute stale.
Background
claw-wall uses a background poller pattern — all agents share one in-memory
conversationStore, and/channel-contextrequests never hit Discord directly. So poll interval is the only knob that controls freshness.The default of 30s was probably conservative, but in practice the trading desk wants 15s channel freshness. The
CLAW_WALL_POLL_INTERVALenv var already exists as the escape hatch, but the default should reflect the intended cadence.Proposed Change
defaultWallPollSecondsfrom30to15incmd/claw-wall/main.goNotes
While investigating a sentinel alert ("weston feed error rate 50.0%"), confirmed that the connection resets in claw-wall logs are transient Discord TCP failures — not caused by poll frequency. Stale Discord context is a separate, lower-severity issue this change would also improve.