[Rahulsharma0810]

Problem

When multiple providers serve the same underlying model, the /v1/models catalog emits identical name fields for every entry.

Example — GPT-5.5 today:

{ "id": "gh/gpt-5.5",          "name": "GPT-5.5" }
{ "id": "cx/gpt-5.5",          "name": "GPT-5.5" }
{ "id": "opencode-zen/gpt-5.5", "name": "GPT-5.5" }
{ "id": "newapi.285743/gpt-5.5", "name": "GPT-5.5" }

Clients that render the name field for display — such as the opencode-omniroute-auth plugin (src/plugin.ts: name: model.name || model.id) — show an identical "GPT-5.5" label for every variant. The user sees the same text four times in their model picker with no way to distinguish which provider will handle the request.

This affects every model served by multiple providers: GPT-4o, Claude Sonnet, Gemini Flash, DeepSeek R1, and many more.

Fix

Add disambiguateCatalogModelNames() in src/lib/modelMetadataRegistry.ts. After the enrichment step, it makes a single O(n) pass over the catalog:

1. Collect all display names and the set of provider prefixes using each name.
2. For names that appear under 2 or more distinct provider prefixes, rewrite name to prefix/BaseName.
3. Names that belong to a single provider are untouched — the clean display is preserved for the common case.

Example — GPT-5.5 after this PR:

{ "id": "gh/gpt-5.5",           "name": "gh/GPT-5.5" }
{ "id": "cx/gpt-5.5",           "name": "cx/GPT-5.5" }
{ "id": "opencode-zen/gpt-5.5",  "name": "opencode-zen/GPT-5.5" }
{ "id": "newapi.285743/gpt-5.5", "name": "newapi.285743/GPT-5.5" }

Single-provider models (e.g. DeepSeek Web-specific models only under ds-web/):

{ "id": "ds-web/deepseek-r1", "name": "DeepSeek R1" }   // unchanged

What is unchanged

• Model IDs, routing, and request parsing are not affected.
• The id, owned_by, root, and all capability fields are not affected.
• Default behavior for single-provider models is identical.
• The change is purely in the cosmetic name field used by display clients.

Why this matters for OpenCode users

The opencode-omniroute-auth plugin (https://github.com/Alph4d0g/opencode-omniroute-auth) renders model names from OmniRoute's catalog name field directly. With 28 providers and 868 total model entries, many models appear under 3–5 different provider prefixes — each showing the same label. There is no way for a user to know which provider they selected from the UI. This PR makes the model picker unambiguous for every affected model.

Files changed

File	Change
src/lib/modelMetadataRegistry.ts	+51 lines: disambiguateCatalogModelNames() helper
src/app/api/v1/models/catalog.ts	2-line wiring (import + pipe enrichedModels through helper)
tests/unit/model-catalog-name-disambiguation.test.ts	6 new unit tests (pure function, no DB)

Tests

node --import tsx/esm --test tests/unit/model-catalog-name-disambiguation.test.ts
# 6/6 pass (pure function tests, no DB needed)

node --import tsx/esm --test tests/unit/models-catalog-route.test.ts tests/unit/models-catalog-low-noise-flag.test.ts tests/unit/model-catalog-name-disambiguation.test.ts
# 52/52 pass

npm run check:file-size
# PASS — catalog.ts stays at frozen baseline 1478

Related

• Closes feat: qualify ambiguous model names with provider prefix in /v1/models #4517
• Follow-up to Feature request: optional low-noise /v1/models output without alias + canonical duplicates #4424 (low-noise catalog mode) and PR feat(models): add opt-in low-noise /v1/models catalog mode #4427 (MODELS_CATALOG_PREFIX_MODE flag).
• Live endpoint measurement: 868 total models, 84 cross-prefix duplicate suffix groups at time of reporting.

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!