[PSA-Team-source]

Summary

Adds UberLLM as a proxy preset, next to the OpenRouter, Requesty and Opper presets. UberLLM is an OpenAI-compatible gateway: base URL https://api.uberllm.dev/v1, GET /models for model discovery, model ids like deepseek/deepseek-v4-flash or anthropic/claude-sonnet-4.6. Prepaid, pay-as-you-go keys; prices are usually below list price and shown per model at https://uberllm.dev/models. Same three files as the Opper preset in #529 (and Requesty in #397). The base URL already ends in /v1, so canonicalBaseUrl keeps it as is.

I work on UberLLM.

Type of change

• Bug fix
• New feature
• Refactor (no behavior change)
• Documentation
• Build / CI / tooling
• Breaking change

Linked issue

None.

Checklist

• I checked the linked issue / relevant context before starting
• pnpm lint && pnpm typecheck && pnpm test passes locally (ran the touched package only, see below; CI will run the full set)
• Added/updated tests for the change
• Added a changeset (pnpm changeset) if user-visible
• Updated docs if behavior changed (not needed)

Tests

• packages/shared: vitest run 23 files, 397 tests passed; tsc --noEmit clean; biome check on the two changed files clean.
• Live check through the project's own code: took the uberllm preset, ran detectWireFromBaseUrl → openai-chat, canonicalBaseUrl → https://api.uberllm.dev/v1, modelsEndpointUrl → GET https://api.uberllm.dev/v1/models returned 200 with 263 models, then complete() from @open-codesign/providers with deepseek/deepseek-v4-flash returned "pong". (Throwaway test file, not committed.)

[github-actions]

Findings

No issues found.

• The new preset is schema-valid: all five fields (id, label, provider, baseUrl, notes) are present in packages/shared/src/proxy-presets.ts:59-65, matching the ProxyPreset schema.
• ProxyPresetId and ProxyPresetIdSchema derive from the PROXY_PRESETS array, so uberllm is automatically included in the enum and covered by the existing "all preset ids are valid ProxyPresetIdSchema values" test (packages/shared/src/proxy-presets.test.ts:18-22).
• Base URL https://api.uberllm.dev/v1 ends in /v1, consistent with the OpenRouter (/api/v1) and Requesty (/v1) presets, so no canonicalization mismatch is expected.
• The id assertion added at packages/shared/src/proxy-presets.test.ts:38 follows the existing test pattern, and the changeset correctly targets @open-codesign/shared as a patch.
• No new dependency, so the ≤30-dep budget and MIT-compatibility constraints are unaffected; BYOK is preserved since this is a user-keyed OpenAI-compatible preset.

Questions

• None.

Summary

Review mode: initial

This is a small, self-contained data addition: one preset entry, one id assertion, one changeset. It is consistent with how the Opper/Requesty presets were added and does not touch provider runtime, persistence, or release/distribution paths. No blocking or non-blocking code issues were found.

Residual risks (not blocking):

• The live behavior of https://api.uberllm.dev/v1 (model listing, complete() auth/wire) was verified only by the author; this run did not independently exercise the endpoint. If the upstream changes host/path, the preset would silently point at a dead endpoint, same as any relay preset.
• There is no explicit assertion pinning the UberLLM baseUrl/provider; a future edit could change them without a test failing. This is optional given the risk profile and existing test style.
• The PR author discloses working on UberLLM; that is a maintainer disclosure decision, not a code finding.

Testing

• Not run (automation). Suggested optional regression test if desired: assert PROXY_PRESETS.find(p => p.id === 'uberllm')?.baseUrl === 'https://api.uberllm.dev/v1', mirroring the official-openai/litellm assertions. Full pnpm test/typecheck will run in CI.

Open-CoDesign Bot