Changelog

CI Status Surfacing on Draft PRs

Every Watari-drafted PR now shows live CI status — the same green/red signal GitHub puts on a PR, mirrored into the Watari dashboard and the Slack notification. No new CI integration: we read the same check_run and status events branch protection already trusts.

• Subscribed to check_run, check_suite, and status webhook events — covers GitHub Actions, Vercel, Cypress Cloud, and legacy CIs like Travis / CircleCI classic
• PR tab shows CI passing, CI running…, or CI failing — N of M checks failed with a per-check breakdown popover
• Slack PR-created notification gets a live CI line that auto-updates via chat.update — one anchor message, no channel flooding
• First-time CI failure fires a separate [ATTENTION] Slack reply with the assigned reviewers @mentioned, so a muted thread still pages the right humans
• Stale-protection: writes are gated on the PR’s current head SHA so force-pushes don’t leave old verdicts behind

Reviewer-Comment Iteration Loop

The first PR is rarely the right PR. Leave a review comment on the Watari-drafted PR — “this fixes the symptom but the bug is in the cache invalidation” — and we propose the follow-up commit on the same branch within minutes. Every iteration learns from the last.

• Watari listens for pull_request_review and issue_comment events on every Watari-drafted PR
• A classifier reads every comment as actionable / question / unrelated / praise — only actionable feedback triggers a regeneration
• An extended-thinking model regenerates the fix against the current branch HEAD and pushes a follow-up commit on the same branch — no force push, no history rewrites
• Per-org cap (default 5 iterations, max 10) with Slack soft-cap notification at iteration 3 if convergence is failing
• Iteration history timeline in the PR tab — each row links to the triggering comment and the resulting commit SHA
• Toggle and cap configurable from Settings → AI; defaults to ON to match the AI-PR-tooling industry default

Feature Request Extraction

• AI now extracts both bugs and feature requests from support tickets
• New item_type field distinguishes bugs from feature requests
• Feature requests create Jira Story issues (bugs remain Bug type)
• Linear issues tagged with Feature Request label for product signals
• Distinct Slack notification formatting for bugs vs feature requests
• Type filter added to the tickets dashboard for quick triage

Watari Launch

• Structured bug extraction from Zendesk and Intercom tickets
• Zendesk and Intercom webhook integrations
• Slack, Linear, and Jira notification delivery
• Full dashboard with bug reports, ticket threads, and analytics
• Razorpay billing with 14-day free trial
• Team management and role-based access