parseModelString returns null for short model ids like "claude-sonnet-5" #234

OS version

macOS (Darwin 25.6.0, arm64)

App version / commit hash

16cc3c8 (upstream/main at time of writing)

Repro steps

import { parseModelString } from './src/shared/utils/modelParser';

parseModelString('claude-sonnet-5'); // → null
parseModelString('claude-opus-5');   // → null

Short model ids of the form claude-{family}-{major} (no minor version, no date suffix) are written into session JSONL logs by some routers/proxies that front the Anthropic API. In the new-format branch the parser requires a 4th part:

if (parts.length < 4) {
  return null;
}

Expected vs actual behavior

Expected: parseModelString('claude-sonnet-5') returns { name: 'sonnet5', family: 'sonnet', majorVersion: 5, minorVersion: null } — the id is a valid model identifier; minor version and date are already optional in every other accepted format.

Actual: returns null, so consumers silently lose model info — model badges show nothing for such sessions and any pricing/cost lookup keyed on parseModelString fails.

Impact

Sessions logged with short model ids lose model attribution across the app (model extraction in EnhancedAIGroup, per-model cost estimates). Happy to send a small PR — the fix is removing the redundant parts.length < 4 guard (2-part input is already rejected above; minorVersion/date parsing already tolerates absence).