feat(providers): provider gaps audit — Lepton AI missing; 8 of 11 already implemented

Description

# Provider audit against 2026-07 catalog Read the delta and produced the following list. Verifying each has a live public docs URL and OpenAI-compat endpoint where applicable. ## Quick wins (free + OpenAI-compat) - **Lepton AI** — https://www.lepton.ai/docs/public_models/model_apis — OpenAI-compat /v1/chat/completions, bearer token, $10 free credit. Near-clone adapter. - **LM Studio** — https://lmstudio.ai/docs/app/api/endpoints/openai — Local OpenAI-compat on localhost:1234. base_url override only. - **Ollama (local)** — https://github.com/ollama/ollama/blob/main/docs/openai.md — Registry has ollama-cloud but not local at localhost:11434/v1. No auth. - **vLLM (self-hosted)** — https://docs.vllm.ai/en/latest/serving/openai_compatible_server.html — Configurable base_url, no auth by default. - **llama.cpp HTTP server** — https://github.com/ggerganov/llama.cpp/tree/master/examples/server — Same trivial adapter shape. - **Text Generation WebUI (oobabooga)** — https://github.com/oobabooga/text-generation-webui/wiki/12-%E2%80%90-OpenAI-API — Trivial base_url adapter. - **Tencent Hunyuan** — https://cloud.tencent.com/document/product/1729/111007 — OpenAI-compat at api.hunyuan.cloud.tencent.com/v1, free hunyuan-lite. Verify overlap with existing 'tencent' entry. ## Moderate (needs adapter) - **Azure OpenAI** — https://learn.microsoft.com/azure/ai-services/openai/reference — OpenAI-shaped but per-deployment URL + api-version query. Paid. - **IBM watsonx.ai** — https://www.ibm.com/products/watsonx-ai/foundation-models — Custom REST + IAM bearer, not OpenAI-shaped. Paid. Skipped: Replicate (prediction-poll, not chat-shaped), Anyscale Endpoints (deprecated/shutting down). Sanity check: README claims 237 total / 90 free, but registry parses to 171 / 83. Worth auditing registry file count vs manifest to confirm authoritative source before quoting 237 publicly. Cross-link: https://github.com/robertpiosik/CodeWebChat — editor-side project with overlapping chatbot-target ambitions; sharing findings. Thanks for maintaining OmniRoute!