Thanks for the matching frontend follow-up — the diagnosis is correct and the JS/Python normalizers should stay in lock-step.

One concern with the proposed fix that mirrors the same issue I just flagged on the backend PR (#1454): split(':').pop() is too aggressive for model IDs that legitimately contain a colon inside the model name. The most common case is OpenRouter's free-tier suffix:

@openrouter:meta-llama/llama-3.1-405b:free

• Old code (substring(s.indexOf(':')+1)) → meta-llama/llama-3.1-405b:free → after / split + -→. → llama.3.1.405b:free ✓
• New code (split(':').pop()) → free → free ✗

That collapses every :free variant to the same normalized key ("free"), which would silently dedupe distinct OpenRouter free models against each other. That is a worse failure mode than the duplicate Primary badge this PR is trying to fix — and it directly conflicts with #1426 (OpenRouter free-tier visibility).

A more conservative shape that handles @custom:provider:model without breaking @provider:model:variant:

if(s.startsWith('@')&&s.includes(':')){
  const parts=s.split(':');
  // @custom:<provider>:<model...> → strip both prefix segments
  if(parts.length>=3 && parts[0]==='@custom'){
    s=parts.slice(2).join(':');
  } else {
    // @<provider>:<model...> → strip only the @provider segment
    s=parts.length===2 ? parts[1] : parts.slice(1).join(':');
  }
}

This needs to land together with the backend fix in #1454 — they have to use the exact same algorithm, or the badge map keys (built server-side) won't match the lookup keys (built client-side). Whichever shape we converge on in #1454, mirror it here verbatim.

Bonus badge label change

The Primary (jingdong) enrichment is reasonable and surgical — I'd keep that part as-is. Two minor notes:

1. Empty-provider safety: m.badge.provider exists in the current map shape, but if a future code path ever sets it to "" you'd render Primary (). Cheap guard: if(m.badge&&m.badge.provider&&m.badge.provider.trim()){...}.
2. The replace(/^custom:/,'').split('/')[0] collapses custom:openrouter → openrouter correctly, but for a raw provider like openrouter/sonoma it gives openrouter, which loses the sub-provider. Probably the right tradeoff for a compact badge — just confirming intent.

Tests

This is renderer-pipeline glue + a normalizer tweak — both should have a targeted Python/JS test:

• tests/test_norm_model_id_javascript.py (or similar) that scrapes _normalizeConfiguredModelKey from static/ui.js and asserts the OpenRouter :free case + the @custom:provider:model case both produce sensible keys.

Labelling bug + frontend + needs-rework. Please rebase against #1454 once that lands and we converge on the algorithm.

cc: @nesquena for the algorithm decision in #1454.