Release v3.1.11 — promote beta to main - #281
Conversation
Merges beta into main for the 3.1.11 release. The conflicts came from #264 (the 3.1.10 gateway-recovery squash landing on main) diverging from the same work as it evolved on beta; every one resolved to the beta side. The resulting tree is byte-identical to origin/beta -- the exact code verified on hardware today. What ships: - codex: route codex turns through the app-server harness (#280). Without agentRuntime, core posts to /backend-api/responses -- a browser endpoint Cloudflare challenges -- and every turn dies with an HTML error page. - codex: restore the credential #278 broke, and migrate legacy auth profiles into the sqlite store core actually reads (#279). - codex: stop mirroring the rotating refresh token into every codex-home, so a single-use refresh token cannot be spent twice (#278). - codex: default ChatGPT auth to gpt-5.5; unblock GPT-5.6 for entitled accounts (#275, #276). - gateway: stop update/boot from bricking the box; parameterised health and recovery waits. - memory: local embeddings so semantic memory works without an API key. - docs/site: point references at clawbox.com; measured Wi-Fi and local-model performance figures. Hardware verification (2026-07-28, boxes .52 and todor): - HTML/Cloudflare failure reproduced on the old build, fixed by the update - 401 profile=- auth failure reproduced, fixed by the update - forced OAuth token rotation x2 on both boxes -- codex survived both - 3.1.10 -> 3.1.11 through the shipped updater: pass - 3.1.5 -> 3.1.11 through the shipped updater incl. reboot: pass, ~4 min - CI: 1499 unit tests, E2E, E2E-install in a real systemd container Known, not a regression (present in 3.1.10, fix queued for 3.1.12): the Codex model picker offers gpt-5.4/gpt-5.4-mini, which free-tier ChatGPT accounts cannot run -- upstream 400s with no failover. The default is gpt-5.5, which every tier runs, so this only bites a user who picks an older model by hand.
|
Warning Review limit reachedYou’ve reached a temporary PR review limit under our Fair Usage Limits Policy. Next review available in: 2 minutes Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. Review details⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: ASSERTIVE Plan: Pro Run ID: 📒 Files selected for processing (55)
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
🦀 ClawReviewFresh PR washed in with the tide — here's the gist. Promotes beta (aeb34cd) to main as the v3.1.11 release, resolving six merge conflicts by taking the beta side throughout — the resulting tree is byte-for-byte what has been running on verified hardware. The headline fixes are: routing Codex turns through the app-server harness so they reach the real Codex API instead of a Cloudflare-challenged browser endpoint; eliminating the double-rotator OAuth bug that burned ChatGPT refresh-token families within hours of a login; and adding a new systemd timer that keeps credential mirrors fresh between reboots. Also ships local embeddings for semantic memory (no API key required) and migrates every clawbox.tech reference to clawbox.com. At a glance
🔎 Heads-up: this looks related to #273 — PR #273 is titled 'Correct Wi-Fi spec and local-model performance in hardware docs'; this PR includes exactly those corrections (Wi-Fi 5 vs Wi-Fi 6, measured tok/s figures, tegrastats methodology note) in clawbox-connect.mdx and requirements.mdx. #273 is likely superseded by or already merged into this release. Good to know
— ClawReview 🦀, scuttling off. General info only — see CodeRabbit for the detailed review. Conventions: docs. |
github-actions
Bot
added
area: ci-e2e
1 participant
Promotes
beta(aeb34cd) tomainfor the v3.1.11 release.The merge conflicted in 6 files because #264 landed the 3.1.10 gateway-recovery work on
mainas a squash while the same work kept evolving onbeta. Every conflict resolved to thebetaside — the resulting tree is byte-identical toorigin/beta, i.e. exactly the code verified on hardware today.What ships
agentRuntime, core posts to/backend-api/responses— a browser endpoint Cloudflare challenges — and every turn dies with an HTML error page.codex-home(fix(codex): stop mirroring the rotating refresh token into every codex-home #278).gpt-5.5; unblock GPT-5.6 for entitled accounts (fix(codex,gemma): unblock GPT-5.6 on ChatGPT auth and stop Gemma provisioning from hanging #275, fix(codex): default ChatGPT auth to gpt-5.5, not gpt-5.4 #276).Hardware verification — 2026-07-28, boxes .52 and todor
401 profile=-auth failure reproduced → updateKnown, not a regression
The Codex model picker offers
gpt-5.4/gpt-5.4-mini, which free-tier ChatGPT accounts cannot run — upstream 400s with no failover. Present in 3.1.10 already; default isgpt-5.5, which every tier runs, so it only bites a user who picks an older model by hand. Fix queued for 3.1.12.