Troubleshooting Playbook

Last reviewed: 2026-03-18 Owner: Support engineering Source of truth:

https://dawnhq.ai/docs/channels
https://dawnhq.ai/docs/integrations
https://dawnhq.ai/docs/getting-started Applies to: first-line support responses Refresh cadence: bi-weekly

Facts

Most setup issues map to one of: credential/auth failure, missing binding, incorrect origin/webhook endpoint, insufficient provider scopes.
First triage sequence:
1. Confirm workspace and active profile.
2. Confirm provider/model availability for that profile.
3. Confirm channel binding is attached and trigger/context mode are correct.
4. Confirm the channel target, webhook, or allowed-origin configuration is correct for that channel.
5. Confirm integration linkage, assigned sources, and read/search access.
6. Re-test with a minimal prompt and known source target.
Missing context often comes from Dawn not having received the earlier discussion; public docs do not promise passive history backfill.
For web widget, CORS/origin misconfiguration is a common root cause.

No response in channel: verify binding attachment and channel credentials first.
No recent discussion/context: verify context mode and whether Dawn was already present to receive those earlier messages.
Source-backed answer missing data: confirm the source is linked and readable.
Web widget issues: validate allowed origins and production API base URL.
Codex not visible: confirm the account linking flow completed and re-open provider/model selection.