mirror of
https://github.com/diegosouzapw/OmniRoute.git
synced 2026-08-13 18:52:18 +03:00
87c53cabac
Ollama is the most popular local runtime, yet OmniRoute only shipped `ollama-cloud` (api-key, hosted) and `ollama-search` (web search) — there was no dedicated card for the local Ollama chat runtime at localhost:11434, so users had to fall back to the generic `openai-compatible-*` provider. Add `ollama-local` to the local catalog: - LOCAL_PROVIDERS entry (default http://localhost:11434/v1, no API key, passthroughModels → lists models via the OpenAI-compatible /v1/models) - classify it as a self-hosted chat provider (SELF_HOSTED_CHAT_PROVIDER_IDS) - route it through the local-baseUrl case group in DefaultExecutor.buildUrl so an empty baseUrl falls back to localhost:11434 instead of OpenAI (same fix class as llama-cpp #3136/#3197) - bump the provider-count slice of PROVIDER_REFERENCE.md (236 → 237) and sync the count in README / AGENTS / CLAUDE Tests: tests/unit/ollama-local-provider.test.ts (catalog shape, local classification, and buildUrl routing — incl. the no-baseUrl OpenAI-fallback regression guard). Refs #5578