fix(browser): reject retired GPT-5.2 aliases - #351
Conversation
|
Codex review: needs maintainer review before merge. Reviewed August 2, 2026, 10:57 PM ET / August 3, 2026, 02:57 UTC. ClawSweeper reviewWhat this changesThe branch rejects retired GPT-5.2 base, Instant, and Thinking aliases before a ChatGPT browser run starts, while retaining API aliases, legacy Pro routing, and the current-model strategy. Merge readinessKeep this owner-authored PR open for normal merge review: current main still attempts to select retired GPT-5.2 browser picker labels, while the proposed early validation is narrow, preserves the API and Pro paths, and directly addresses the linked open report. Priority: P2 Review scores
Verification
How this fits togetherOracle resolves a CLI model request into either an API model or a ChatGPT browser picker target. Browser configuration is built before Chrome is launched, so this change can replace a failed picker lookup with an immediate, actionable CLI error. flowchart LR
A[CLI model request] --> B[Engine and model resolution]
B --> C[Browser configuration]
C --> D{Picker target available?}
D -->|Retired GPT-5.2 alias| E[Actionable CLI error]
D -->|Supported target or current strategy| F[Launch Chrome]
F --> G[ChatGPT model picker]
Before merge
Agent review detailsSecurityNone. Review metrics
Root-cause clusterRelationship: Members:
Proposal only: this assessment does not dispatch repair, suppress jobs, mutate sibling items, close, or merge anything. Merge-risk optionsMaintainer options:
Technical reviewBest possible solution: Merge the early browser-only availability gate after confirming the failing build checks are unrelated, so retired ChatGPT picker aliases receive a deterministic error while API and legacy-Pro compatibility remain intact. Do we have a high-confidence way to reproduce the issue? Yes—current main maps the affected browser aliases to GPT-5.2 picker labels and continues into browser configuration, while the linked report records those labels as absent from the live picker. This review did not launch a browser because the checkout is read-only. Is this the best way to solve the issue? Yes—the browser-configuration gate is the narrowest maintainable fix because it runs before Chrome startup, applies to CLI and TUI callers, and leaves API aliases, legacy Pro routing, and explicit current-model selection unchanged. AGENTS.md: found and applied where relevant. Codex review notes: model internal, reasoning high; reviewed against 666e01879a80. LabelsLabel changes:
Label justifications:
EvidenceWhat I checked:
Likely related people:
Rank-up movesOptional improvements that raise the rating; they are not merge blockers.
Rating scale
Overall follows the weaker of proof and patch quality. Workflow
|
|
ClawSweeper status: review started. I am starting a fresh review of this pull request: fix(browser): reject retired GPT-5.2 aliases This is item 1/1 in the current shard. Shard 0/1. This placeholder means the worker is alive and reading the current context. I will edit this same comment with the actual review when the claws are done clicking. Crustacean status: shell secured, claws on keyboard, evidence pebbles being sorted. |
|
Landed as Proof:
The first CI run exposed stale GPT-5.1 browser fixtures in the full suite; those were updated to current GPT-5.4 fixtures, the full local gate was rerun, and the exact-head CI rerun passed everywhere. |
Summary
currentbrowser selectionFixes #344.
Proof
pnpm exec vitest run tests/cli/browserConfig.test.ts tests/cli/runOracle/runOracle.request-payload.test.ts— 105 passedpnpm run typecheckpnpm run docs:check— 76 flags, 6 filespnpm run buildgpt-5.2,gpt-5.2-instant, andgpt-5.2-thinkingexits 1 with the retirement message before Chrome startupgpt-5.2-probrowser target resolves to Pro; APIgpt-5.2, browsergpt-5.5, and--browser-model-strategy currentall exit 0