Operations — bugs and support

Status: draft scaffold. Most of this is tribal today. Document the actual flow as we observe it for two weeks, then codify.

PH analogue: PostHog folds this into per-team rotation docs + their support team handbook. No single canonical "ops" page — we'd be inventing the shape.

Support flow (Chatwoot)

TODO. Things to capture:

• Inbound channels (in-app, email, social).
• Chatwoot routing and labels.
• SLA policy (memory: SLAs assigned by support-bot's apply_sla_policy call, not by a Chatwoot automation rule) — link to internal/support-bot/ once that's written up.
• Escalation path: support → eng on-call.
• Templates / canned responses.

Bug intake

TODO.

• Where bugs come from (Chatwoot, Linear, Sentry, internal Slack).
• Triage cadence — who, when, how often.
• Severity rubric (S0/S1/S2/S3 — define).
• "Reproduce first" vs "ticket first" rule.

On-call

TODO.

• Rotation (who's on, for how long).
• Pager source (Sentry? Better Stack? PostHog alerts?).
• Hours of coverage (business hours? 24/7?).
• Hand-off ritual.

Bug-to-fix lifecycle

TODO.

• Linear states: Triage → Backlog → In Progress → In Review → Done.
• When to write a postmortem (see docs/postmortems/README.md).
• Customer comms loop — who tells the reporter the bug is fixed.

Tooling map

• Support inbox: Chatwoot (dashboard)
• Bug tracker: Linear (project studyflash)
• Errors: Sentry
• Replays: PostHog session replay
• AI traces: PostHog LLM analytics
• Support assistant: internal/support-bot/ (CLAUDE.md inside)

TODO

• Document the support-bot decision tree.
• Document who has Chatwoot admin vs agent access.
• Add a "first hour of a customer-reported outage" runbook.