release: v1.22.1 — ship the QUA-1580 cloud MCP routing fix - #160
Conversation
The QUA-1580 fix merged on 2026-07-28 but has never been released. The newest tag, v1.22.0, points at 29a5fd2 — the commit immediately before it — so `main` has been exactly two commits ahead of every published build, and those two commits are the fix. Because the installer resolves `releases/latest`, every user installing qmax-code since then has received a binary whose `serve --mcp` still calls the REST API directly: tool calls set no trace context (no session history at all) and return UI-shaped payloads carrying every script's full source, which no server-side response budget can reach. Verified against the tags: none of v1.20.8, v1.20.9, v1.21.0, v1.21.1, v1.21.2 or v1.22.0 contains 8f816d2. Changelog entry + version constant only; no functional change. CI is green on 56a5940, the commit this is based on. Refs QUA-1725, QUA-1580.
Sigilix OverviewEffort: 1/5 (trivial) Quality gates
Summary — latest pushBumps the version constant to 1.22.1 and adds a changelog entry to formally release the previously merged but unshipped QUA-1580 cloud MCP routing fix. The functional changes (routing authenticated tool calls through the cloud MCP endpoint and minting MCP-compatible tokens after browser login) are already on main; this PR exists solely to cut the release tag so the installer distributes the fix. The specialist review verdict is a clean APPROVE with zero findings. Important files
Confidence: 5/5Purely mechanical version bump and changelog addition with no functional code changes; CI is green on the base commit and the specialist review returned zero findings.
|
QualityMax ReviewVerdict: COMMENT · Confidence: insufficient — 139 model output tokens (minimum 200) Files eligible: 2 · Files reviewed: 2 · Files with findings: 0 · Findings: 0 · Inline cards: 0 Priority findings
Review gates
Important files
Review lifecycleUse the inline cards to inspect evidence and suggested remediation. Re-run the QualityMax review after pushing a fix; unchanged cards are identified by their stable finding marker. Dismiss with a reason through the existing QualityMax/GitHub review feedback flow. 0 prior card(s) are stale/resolved on this head. Proof legend: VERIFIED independently judged patch · REPRODUCED verified finding · GROUNDED deterministic evidence · MODEL-ONLY model judgment. QualityMax project results are available in the configured project. Receipt · commit |
|
| Gate | Result |
|---|---|
| 🔍 AI diff review | ✅ Clean · gemini-3.1-flash-lite · completed · 1 eligible / 1 reviewed · gemini-3.1-flash-lite |
| 🔍 SAST | completed · 2 eligible / 2 reviewed · gemini-3.1-pro-preview |
| 🔍 Canonical PR review delivery | completed · 0 eligible / 0 reviewed · exact-head review #4839852033 and overview #5160616806 confirmed |
| 🧪 Repo Tests | ✅ 553/553 passed (go) |
| 🤖 AI Tests |
Powered by QualityMax — AI-Powered Test Automation
Refs QUA-1725, QUA-1580.
The problem
The QUA-1580 fix merged on 2026-07-28 and has never been released.
mainis exactly two commits ahead of the newest tag, and those two commits are the fix. Checked every recent tag for8f816d2:The installer resolves
releases/latest, so every user who installed qmax-code since Jul 28 got a binary without the fix. Theirserve --mcpstill calls REST directly, which means:list_scriptscall, and no server-side response budget can reach it — the platform-side work in QUA-1724 only covers/api/mcp/.QUA-1580 was marked Done on a merged PR. Merged ≠ shipped.
This PR
Changelog entry + version constant. No functional change — the fix itself is already on
main.CHANGELOG.md— new## [1.22.1]section. Both commits landed with no changelog entry, so[Unreleased]was empty despite the fix sitting there.main.go—Version1.22.0 → 1.22.1. Cosmetic for released builds (release.ymlstamps the tag via ldflags) but correct forgo buildfrom source.Patch, not minor: bug fix plus the token minting it needs, no new surface.
Verification
origin/main, 0 ahead / 0 behind at time of branching; both fix commits confirmed present.56a5940— the exact base commit — for bothCIandQualityMax Go Tests.go vet ./...clean locally. Localgo test ./...showsinternal/mcpok; five packages (setup,skills,sysutil,tui,vnc) hit the 11m budget on this machine for want of a TTY/network, which is why the CI result above is the load-bearing signal rather than my laptop.After merge
Tag
v1.22.1on the merge commit and push —release.ymlfires onv*and handles tests, six-platform builds, archives, and the GitHub release. I have not tagged anything; that's the deliberate manual step.Loose end, not fixed here
Desperado/update-qmax-docs-orch-modecarries a commit7601b2b test+release: … release v1.22.0that differs from the actualv1.22.0tag (29a5fd2). If that branch merges as-is you'd get a second, conflicting 1.22.0. Worth untangling separately.