chore(release): v0.12.11
Issue for this PR
Closes #1749
Release PR. It promotes the v0.12.10..main payload (17 PRs) to latest as v0.12.11. Highlights: #1706 "feat: Claude subscription support through a Claude Agent provider"; workspace review and roles (#1685, #1707, #1708, #1716); #1626 "feat: workspace rules in Claude Code and Codex sessions through the plugins"; and #1710 "feat: column-level impact answer in impact_analysis".
Type of change
- Bug fix
- New feature
- Refactor / code improvement
- Documentation
What does this PR do?
Adds the 0.12.11 changelog entry, the release adversarial tests (packages/opencode/test/skill/release-v0.12.11-adversarial.test.ts, 17 cases), and the fixes from the release review.
Review fixes in this PR. They come from five persona reviews of the real diffs (CTO, Product Manager, End User, Tech Lead, and a support engineer on upgrade day as the Chaos Gremlin): 5 SHIP WITH NOTES, 0 HOLD, 0 P0.
- Upgrade from the removed Claude Pro/Max sign-in. I reproduced it: a 0.12.10 profile with only that login and
--model anthropic/claude-sonnet-4-6failed with a bare "Model not found".run, the TUI andservenow add "The Anthropic provider needs an API key. For a Claude Pro or Max plan, run /connect and pick Claude subscription." The hint appears only when nothing is suggested, so a model-name typo is unaffected. Without--model, the run already fell back to the default model. (altimate/provider/model-hint.ts,session/prompt.ts,cli/error.ts) - Claude subscription
/connectfailure. After the 5-minute wait the toast now says to runclaude auth login(Claude Code 2.1.296 or later) instead of "OAuth authorization failed". A Claude Code start that times out names the network or proxy. (dialog-provider.tsx,claude/probe.ts,claude/sessions.ts) impact_analysis. A column answer that could not be built is said in the report, so it no longer reads as "no column impact". (altimate/tools/impact-analysis.ts)/rules. Removing a request from the workspace stops serving it to the whole team, so it now asks for confirmation, opening on No. (plugin/tui/altimate/rules.tsx)skill publish. A member's publish says approval happens in the Altimate web app.- Docs.
- New
impact_analysisreference in the lineage tools page. tool report-event.- Claude subscription notes in the providers page: an
ANTHROPIC_API_KEYin the shell is used instead of the subscription, proxy settings, the Windowsclaude.exepath, and deleted chats. The duplicate "Claude Pro/Max" row is renamed.
- New
Deferred, with issues:
- #1744 "Claude subscription follow-ups": the Windows shim, the failure reason through the plugin API, a startup notice outside the TUI, the environment passed to
claude, and an RPC cancel test. - #1745 "the subscription detection route answers any local process".
- #1746 "background subagent records are not removed when a session is deleted".
- #1747 "five CLI suites never run in CI".
- #1748 "workspace review follow-ups".
The picker change that moves API-key rows down is a deliberate part of #1706, and the changelog states it.
How did you verify your code works?
- Preflight:
bun script/release-preflight.ts --version 0.12.11 --stage pregave 9 PASS, 0 WARN, 0 FAIL on a clean tree atorigin/main9e8294ec5e. The release notes will compare against v0.12.10. - Reviews: five persona reviews read every diff.
- Static checks:
bun turbo typecheckpassed on all 13 packages, and the marker guard is clean. - New tests:
- Release adversarial tests, 17 pass. They cover subscription detection against hostile login files (a prototype-pollution key, non-object JSON, a directory, a 2 MB file, a symlink), the removed-login flag, the model hint as a truth table, the storage key guard, and source checks.
- A new
cli/errortest, and animpact_analysisassertion.
- Upgrade reproduction: a 0.12.10-style profile with only an Anthropic OAuth login, before and after the fix.
- Full suite on
26540944f9, split the way CI runs it:- 4 shards, about 22,100 tests: all pass except 6 MCP tests (
mcp.headers,mcp HttpApi status), which fail the same way on plainmainon that machine because of its local MCP config. learn/cli: 73 pass. Subprocess pass: 40 pass. Other CLI suites: 114 pass. TUI: 364 pass.
- 4 shards, about 22,100 tests: all pass except 6 MCP tests (
- Release binary smoke test:
OPENCODE_VERSION=0.12.11 bun run pre-releasereports "ALL CHECKS PASSED", and bothaltimateandaltimate-codereport0.12.11.
Cost side effects
None. The review fixes change messages, docs and one confirmation dialog. The Claude subscription provider (#1706, already on main) uses the user's own Claude plan or key.
Screenshots / recordings
Not included. The UI changes are a toast message and a confirmation dialog.
Checklist
- I have tested my changes locally
- I have not included unrelated changes in this PR
- I checked for cost side effects (model calls, queries or cloud resources that grow with user data) and described them, or wrote "none"
🤖 Generated with Claude Code
https://claude.ai/code/session_01XM3J9mB75wZRnzTSGcW2kc
Summary by CodeRabbit
- New Features
- Added guidance for connecting a Claude subscription when the Anthropic provider is unavailable.
- Documented dbt model- and column-level impact analysis, including affected resources and lineage results.
- Changed
- Workspace skill requests now require confirmation before removal; withdrawal remains available directly.
- Claude connection and startup errors now provide more specific troubleshooting guidance.
- Bug Fixes
- Impact analysis reports when a column-level answer is unavailable without repeating an already-reported failure.
- Documentation
- Clarified Claude setup, configuration options, chat deletion behavior, and plugin event reporting.