feat: add Opper as a model provider preset

[Felixkw12]

Opper is an EU-hosted AI gateway run by Opper Technology AB (Stockholm): one API key and one OpenAI-compatible endpoint reach models from 30+ upstream providers at each provider's own token rates, with no markup, and requests can be pinned to EU-hosted regions.

It is wired exactly like the Vercel AI Gateway preset: an OpenAI chat completions endpoint with an empty model list, so the catalog is imported on demand from GET /models.

• src/shared/model-provider-preset-catalog-extended.ts: the preset
• src/shared/model-provider-preset-types.ts: the preset id
• packages/provider-catalog/src/index.ts: the shared catalog entry
• src/main/models-dev-catalog.ts: Opper is published on models.dev as opper, so the preset maps straight onto that catalog
• docs/model-provider-presets.md: the Built-in Providers entry

No icon asset, matching Vercel AI Gateway, so the diff stays additive. Model ids come in two forms and the docs entry explains both: a bare pool name such as claude-sonnet-4-6 routes across every provider serving that model, and provider/model such as aws/claude-sonnet-4-6-eu pins one provider or region.

I work at Opper. Happy to adjust anything to match project conventions.

Not verified locally: npm test needs the full dependency install, which I did not run here. Every change is a data entry mirroring the adjacent gateway preset, plus one union member and one catalog mapping.

Summary / 概要

Why / 背景

Changes / 变更

Media / 截图或录屏

• UI changes: attach a video or GIF here.
• 界面变更：请在这里附上视频或 GIF。

Tests / 测试

• Logic changes: list unit tests added or updated.
• 逻辑变更：请列出新增或更新的单元测试。

Validation / 验证

• I agree that this contribution is submitted under the Contributor License Agreement. / 我同意本贡献遵循 Contributor License Agreement 提交。
• npm run test
• npm run typecheck
• npm run build
• npm run dev (if runtime or UI behavior changed / 如果运行时或界面行为有变化)
• UI change: video or GIF attached / 界面变更：已附上视频或 GIF
• Logic change: unit tests added or updated / 逻辑变更：已新增或更新单元测试

Notes / 备注

[XingYu-Zhong]

Thanks for the contribution — the preset is wired correctly and opper on models.dev checks out (api: https://api.opper.ai/v3/compat, matching the preset base URL).

One gap before this can merge: UNAMBIGUOUS_URL_MATCHES in src/main/models-dev-catalog.ts is missing 'https://api.opper.ai/v3/compat': 'opper'. Profiles whose provider id is not exactly opper — e.g. a duplicated opper-2 account or a custom provider pointing at the Opper endpoint — would fail to resolve the models.dev catalog and silently get no model metadata. The Vercel AI Gateway preset covers both the profile-id path and the URL path; Opper should match it.

Since "Allow edits from maintainers" is off on the fork branch, I've carried your commits into #1340 and added the fix + test cases there. This PR is superseded by #1340 — happy to credit you in the merge.

[XingYu-Zhong]

This PR has been superseded by #1340, which has been merged into develop.

#1340 carries your two original commits (authorship preserved) plus one fix on top:

• fix(providers): resolve Opper base URL to the models.dev catalog — adds https://api.opper.ai/v3/compat to UNAMBIGUOUS_URL_MATCHES, so custom/copied profiles pointing at the Opper URL also resolve the models.dev catalog, plus test coverage.

It also picked up baseline fixes from develop (#1341) that were blocking CI. Thanks for the contribution!

[Felixkw12]

Thanks for catching the UNAMBIGUOUS_URL_MATCHES gap, that was a real hole. A profile id of opper-2 or a custom profile pointing at the same endpoint would have resolved nothing and silently lost model metadata.

Closing in favour of #1340. Appreciate you carrying the commits over and adding the test cases.