fix(cursor): support current Grok 4.5 Fast wire IDs

[jbaehova]

Summary

• recognize Cursor's current Grok 4.5 Fast wire IDs, where -fast follows the effort tier
• align Grok 4.5 and Grok 4.5 Fast with Cursor's current low / medium / high effort names
• use one wire-ID composer for live discovery and outgoing Cursor requests
• keep discovery compatibility with the older grok-4.5-fast-{effort} ordering

Why

Cursor now exposes Fast variants as IDs such as cursor-grok-4.5-high-fast, while the adapter only matched and emitted grok-4.5-fast-high. Live model filtering therefore removed cursor/grok-4.5-fast from the GUI and Codex catalog even when the account could use it, and direct requests selected a wire ID Cursor no longer serves.

Cursor staff documented the current slug migration from grok-4.5-fast-{medium,high,xhigh} to cursor-grok-4.5-{low,medium,high}-fast: https://forum.cursor.com/t/cursor-grok-4-5-high-fast-doesnt-offer-a-50-discount-at-all/165551/8

Related context: #731 and the closed wrong-target PR #730.

User impact

Accounts whose live model list includes Grok 4.5 Fast can retain cursor/grok-4.5-fast as a separately selectable model. Requests now send the matching trailing-Fast wire ID, including the default/top-tier grok-4.5-high-fast.

Test plan

• bun test tests/cursor-discovery.test.ts tests/cursor-effort-suffix.test.ts
• bun run typecheck
• bun run prepush

Summary by CodeRabbit

• New Features

  • Added support for low, medium, and high effort levels across additional Cursor models, including Grok and Fast variants.
  • Improved recognition of Cursor model identifiers so supported standard and Fast models appear correctly.

• Bug Fixes

  • Fixed model availability matching for current and legacy effort identifier formats.
  • Prevented mismatched model and effort combinations from being selected.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: e346ca73-4be3-4ac4-aaa7-fee5c64eac6b

📥 Commits

Reviewing files that changed from the base of the PR and between afe2aa3 and c879ec9.

📒 Files selected for processing (5)

• src/adapters/cursor/discovery.ts
• src/adapters/cursor/effort-map.ts
• src/adapters/cursor/request-builder.ts
• tests/cursor-discovery.test.ts
• tests/cursor-effort-suffix.test.ts

---

📝 Walkthrough

Walkthrough

Cursor effort mapping now supports Grok low/medium/high tiers and places -fast after the effort suffix in wire IDs. Request normalization and account discovery use the new format while retaining legacy matching.

Changes

Cursor effort wire-ID handling

Layer / File(s)	Summary
Effort mapping and request construction src/adapters/cursor/effort-map.ts, src/adapters/cursor/request-builder.ts, tests/cursor-effort-suffix.test.ts	Grok tiers now use low/medium/high. cursorWireModelIdWithEffort builds standard and -fast wire IDs. normalizeCursorModelId uses the helper. Tests cover clamping, canonical ladders, and Fast identifiers.
Account discovery compatibility src/adapters/cursor/discovery.ts, tests/cursor-discovery.test.ts	Account matching recognizes standard IDs, current -{effort}-fast IDs, and the legacy -fast-{effort} format. Tests reject mismatched combinations and retain both Grok variants during filtering.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Possibly related PRs

• lidge-jun/opencodex#730: Modifies Cursor effort-ID construction and discovery for Grok Fast wire IDs.
• lidge-jun/opencodex#318: Modifies Cursor model-ID handling in discovery and request building.
• lidge-jun/opencodex#646: Modifies Cursor effort-tier mapping and discovery tests.

Suggested reviewers: wibias, lidge-jun, koopmannleon19977-cmyk, ingwannu

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly identifies the Cursor fix for current Grok 4.5 Fast wire IDs, which is the primary change in the pull request.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands.}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c879ec95ce

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Preserve the discovered Fast wire-ID ordering

When GetUsableModels returns only a legacy ID such as cursor-grok-4.5-fast-medium, isCursorModelAvailableForAccount deliberately keeps grok-4.5-fast visible, but every request is now unconditionally rewritten to grok-4.5-medium-fast. For the exact legacy-account context this compatibility branch handles, the selected model can therefore fail with ERROR_BAD_MODEL_NAME; either stop exposing legacy-only discoveries or retain the discovered ordering when constructing the request.

Useful? React with 👍 / 👎.