peerd-preview-v0.1.4
Goal-mode hardening, side-panel state fixes, an end-to-end test tier, and
provider default-model selection. All changes were verified green and
reviewed by adversarial-swarm passes before merge.
Added
- Live-extension E2E tier. A reusable raw-CDP harness with goal /
stop / error scenarios. It runs the real chassis with the model faked
at the wire, and it is wired into CI (#57). It also adds a local,
npm-free visual-regression layer: self-contained PNG decode and pixel
diff against committed baselines, deliberately kept out of blocking
CI (#64). - Provider-aware default models and WebVM terminal fixes (#62).
Fixed
- Goal-mode autonomous loop. Durable Stop now reaches a
vault-lock-paused run. Resume sequencing and cap-boundary correctness
are fixed (#55). The Goal bar and Stop rehydrate when a surface
connects or reconnects mid-run (#59). Goal-resume is ordered before
auto-resume on interactive unlock, and durable-Stop is now awaited on
steer, new-chat, and archive (#63). - Spend-limit halt banner now persists across unrelated state pushes
(Plan/Act toggle,/system,/tools, settings) instead of vanishing
mid-halt (#54).
Changed
- README embeds the demo video after the intro (#65).
Install (preview): Firefox: click the .xpi on this page. Chrome: drag the .crx into chrome://extensions with Developer mode on. Most users want the store packages (see the README). The auto-update feeds are attached here and served at peerd.ai/updates/.