Repository navigation
v0.61.1: watch Auto mode work
Auto mode now shows its work. Start an Auto chat in the console to watch Ouroboros interview itself, write the spec, review it, run it and verify it, step by step. You see a live story of the run instead of a terminal log.
🎬 A live view for Auto mode
Before, an Auto chat showed the raw ouroboros auto log in a dark box. Now it shows (#3028, #3037):
- A six-step stepper: 인터뷰 → 스펙 작성 → 스펙 검토 → 실행 → 검증 → 결과 (Interview → Write spec → Review → Run → Verify → Result). Each step shows a short caption ("질문 6", "A등급", "기준 2/2", "파일 2개 변경"). Repairs show as a side state on the step they belong to.
- The automatic interview, as a conversation. Each question comes with the answer Ouroboros picked and a badge for where that answer came from: from the code, from your goal, existing convention, default, assumption, out of scope.
- A small ambiguity sparkline shows the interview converging.
- When the interview ends, it folds into one line ("질문 13개에 자동으로 답했어요") that you can expand.
- The spec, the review and the run, in place.
- The Seed card carries the review grade and its history.
- The run card fills in acceptance-criterion tiles as they pass.
- The result card lists the files that changed in your folder.
- Clear endings. Stopped, failed, unverified and nothing-ran each get their own wording and a next step, never a bare exit code.
- The raw log is still there, folded under "자세한 로그 보기".
- Light and dark, phone widths, five languages, screen-reader labels and reduced motion. It went through an independent design review before release.
- For scripts:
ouroboros autohas a new--progress-jsonl PATHoption that writes every progress event as one JSON line.
🛠 Fixes
- Auto Seeds keep only your requirements (#3036). After a Seed review repair, internal pipeline instructions could end up in the Seed's constraints, for example "Resolve Seed QA feedback before execution…". Repairs now add only real product constraints and your confirmed non-goals.
- Telemetry counts each day once (#3032). Under contention, two processes could both send the same day's usage rollup. PostHog deduplicated it, but the lease clock is now read under the lock.
- Test reliability (#3030). Fixes a test that failed on Python 3.14 because of a changed error message, and a test that failed when a hash happened to contain its marker.
Install or upgrade in one line
# macOS / Linux / WSL 2
curl -fsSL https://raw.githubusercontent.com/Q00/ouroboros/main/scripts/install.sh | bash# Windows (PowerShell 5.1+ or pwsh 7+)
irm https://raw.githubusercontent.com/Q00/ouroboros/main/scripts/install.ps1 | iexAlready installed? Run ouroboros update, or ooo update inside your agent. Then open the console with ouroboros gui and start an Auto chat.
What's Changed
- test: stop two tests failing on Python 3.14 wording and digest luck by @Q00 in #3030
- fix(telemetry): read the rollup lease clock under the lock by @Q00 in #3032
- feat(gui): show Auto chats as a live structured progress view by @Q00 in #3028
- fix(auto): keep Seed QA repair instructions out of Seed constraints by @Q00 in #3036
- fix(gui): polish the Auto chat view from the design review by @Q00 in #3037
- chore(release): prepare v0.61.1 by @Q00 in #3038
Full Changelog: v0.61.0...v0.61.1