[PSA-Team-source]

Summary

• Add a scoped community configuration example for UberLLM under docs/config_example/.
• Link it from docs/config_example/README.md.
• Use provider-level modelsDiscovery with https://api.uberllm.dev/v1 and modelInfoFormat: "openrouter", with no static model list or filters.

The one thing that is specific to this provider: UberLLM's /v1/models returns OpenRouter-style inline fields (context_length, top_provider.max_completion_tokens, architecture modalities, supported_parameters, pricing.prompt/completion/input_cache_*), so the existing openrouter enricher fills limits, modalities, tool support and cost from the same response with no extra request. The notes say how to drop it if someone does not want per-model costs set.

Mirrors #49.

Validation

• Parsed the example's JSON block (comments stripped) and fed it through parseProviderDiscoveryOptions + discoverInventory from src/v2 against the live API: 263 models discovered; deepseek/deepseek-v4.1-flash came back with context 1048576, output 384000, tools true, input text,image, and cost set.
• npx vitest run test/openrouter-model-info.test.ts (2 passed)
• git diff --check
• Tested with deepseek/deepseek-v4.1-flash: a chat completion on the discovered model id returned "pong" (HTTP 200).

Current per-model prices (usually below list price): https://uberllm.dev/models

Disclosure

I work on UberLLM. This contribution is limited to the documented community-example scope.