models: name GPT-5.6, Grok 4.5 and ClinePass slugs instead of showing raw ids

[therickfactr]

Summary

• Adds SHORT_NAMES entries for model ids that price correctly but had no display entry, so the By Model panel rendered a raw slug next to properly named siblings: gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna, grok-4.5, qwen3.7-max, minimax-m3, mimo-v2.5-pro.
• Display-only. No dollar amounts move — pricing already resolved for all of these (mimo-v2.5-pro is unpriced upstream both before and after; this only gives it a name).

Before / after, from a real local report:

qwen3.7-max      $0.562      ->  Qwen 3.7 Max     $0.562
minimax-m3       $0.154      ->  MiniMax M3       $0.154
mimo-v2.5-pro    $0.0090     ->  MiMo v2.5 Pro    $0.0090
grok-4.5        ~$10.82      ->  Grok 4.5        ~$10.82
gpt-5.6-sol      $280.56     ->  GPT-5.6 Sol      $280.56
gpt-5.6-terra    $122.98     ->  GPT-5.6 Terra    $122.98

Notes on the non-obvious ones

GPT-5.6 variants are listed individually, not as a bare gpt-5.6. A base entry would swallow every future gpt-5.6-* via the prefix match and hide the variant behind a sibling's label — precisely what getShortModelName's version-boundary rule exists to prevent. An unlisted variant still falls through to its raw id, and a test pins that.

grok-4.5 is the model, not the harness. The Grok Build CLI reports the model it runs as current_model_id, so the id takes the model's own name. Ids that really are grok-build* keep the "Grok Build" label — also covered by a test, since that distinction is easy to regress.

ClinePass needed no new prefix handling. Its ids arrive as cline-pass/<slug>; getShortModelName's path fallback already strips the prefix and re-resolves the bare slug, exactly as it does for accounts/fireworks/models/<slug>. Adding a dedicated prefix rule would have been dead code. Tests cover the cline-pass/ path so the behavior stays intentional rather than incidental.

MiniMax M3 is mapped under both spellings. The gap-filled OpenRouter key is the lowercase slug while sessions report the capitalized name; SHORT_NAMES matching is case-sensitive, and the case-insensitive index covers pricing only.

Testing

• I have tested this locally against real data (not just unit tests)
• npm test passes
• npm run build succeeds

Every id above was taken from real local sessions, not from documentation — gpt-5.6-luna in particular only surfaced by grepping actual Codex transcripts (7 records), which is why it is included alongside the two obvious variants.

npm test: 20 failed / 2619 passed against a main baseline of 23 failed / 2612 passed on the same machine and commit base. No file fails here that does not also fail on main; the failures are the pre-existing cache-refresh-lock / parser durable-orphan / app/electron/cli set plus CLI-subprocess tests that flake under parallel load and pass in isolation. The 26 errors in both runs are app/renderer tests needing jsdom from app/'s own dependency tree.

Five new assertions in tests/models.test.ts cover the variant-collapse guard, the Grok Build distinction, the cline-pass/ path fallback, and both MiniMax spellings.

[ozymandiashh]

Maintainer note: the check failure was the Co-Authored-By: Claude ... trailer, which this repo blocks in commit messages. I amended the commit in place to drop that one line (diff, authorship and the rest of the message untouched) and force-pushed, so remember to git pull --rebase before any further local work on this branch. Review of the substance follows separately.

[ozymandiashh]

Verified locally (CI here does not run tests): 150/150 in models.test.ts, tsc clean. Display-only, each non-obvious mapping decision is argued in a comment and pinned by a test - the variant-collapse guard for gpt-5.6-*, the model-vs-harness split for grok-4.5, the path fallback for cline-pass/ slugs and the case-duplicate for MiniMax. Merging.