🤖 refactor: delete the auto routing keys

[ibetitsmike]

Summary

Workspace creation no longer copies the creation composer's Auto choices into the new workspace's autoModelRouting: and autoThinkingLevel: keys or its per-agent routing record (syncCreationPreferences). The first send already carries the composer's Auto flags, read through the resolver from #5909, and the backend saves them for the agent (#5894). With that copy gone nothing reads autoModelRouting:{scope}, autoThinkingLevel:{scope} or autoRoutingChoiceByAgent:{workspace}, so the key helpers, their storage registry entries, recordAutoRoutingChoiceForAgent and the autoRoutingEnabled option of useCreationWorkspace are deleted.

Startup cleanup (removeDroppedCacheKeys) now deletes leftover autoModelRouting:, autoThinkingLevel: and autoRoutingChoiceByAgent: values from older builds.

Stack

Milestone M4 of the server-side settings stack, on top of M3 (#5869 to #5879): per-workspace AI picks (model, thinking level, reasoning mode, agent and Auto routing) are stored in workspace metadata in config.json instead of per-browser localStorage keys, so every browser shows the same picks. An unsent picker change stays in memory in its tab until a send saves it, and a reload before sending drops it, as the plan accepts. Layer 18 of 20.

Testing

• "the first send carries the creation composer's Auto picks to the new workspace" (useCreationWorkspace.test.tsx): with the model's Auto from the agent's configured default and thinking Auto picked in the creation composer, the first send carries both flags. It replaces "records only creation routing picks that differ from the agent's Auto default (experiment %p)", which tested the removed copy.
• Deleted "keeps the newest per-agent routing choices on disk as agents accumulate" (modelChange.test.ts) with the routing record it covered; persistedStateBudget.test.ts drops the removed keys.
• "removes only keys with a dropped cache prefix" (legacyLocalStorageCleanup.test.ts) covers the removed prefixes.
• make static-check and the touched test suites pass at the top of the stack; CI checks each layer on its own.

Review record

• Remote UAT through Coder Agents: Round 1: FAIL on efe7e29096, because a fork without a message dropped the source's unsent picks (fixed by 🤖 feat: fork a workspace with the source composer's unsent AI picks #5897, 🤖 refactor: resolve the workspace agent from metadata and an unsent pick #5908 and 🤖 refactor: resolve auto routing from the unsent pick, saved flags and agent defaults #5909); everything else passed. Round 2: PASS on 80f2b5379e: forks from all five entry points keep the unsent picks, startup removes the retired keys, and the round 1 checks still pass; the plan handoff was not exercised because the mock model cannot propose a plan. The Codex round 1 fixes made since then have no remote UAT yet; that is left to the next UAT round on 🤖 refactor: keep preferences and workspace AI picks on the server #5916.
• Codex round 1: thumbs-up on 8e8ffc3daf, no findings. Mike asked to land the stack as one PR, so further review runs on 🤖 refactor: keep preferences and workspace AI picks on the server #5916.
• M4 test audit: no gap found in this layer (audit at efe7e29096, probe edits in production code).

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-09T00:26:45.312353Z	73ae413	New commits
🔒 Security Review	✅ Completed 2026-10-09T00:26:33.919877Z	73ae413	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[ibetitsmike]

Landed in #5916 (7b9899d), which collapsed this stack into one squash merge.