0.0.81

Otoroshi LLM Extension — v0.0.81

This release makes the gateway self-describing — clients can now discover which providers and model types are available at runtime — and significantly broadens OpenAI-compatible and OVH AI Endpoints coverage across embeddings, images, audio and moderation.

✨ Highlights

Discover providers & capabilities at runtime

Two new read-only endpoints let any client (or your own UI) find out what the gateway can do, without hard-coding a list.

• GET /providers — the catalog of every provider the gateway can talk to, each with the capabilities it exposes (text, audio, image, ocr, embedding, moderation, video).
• GET /model-capabilities — the list of model types (modalities) the gateway supports, each with the providers that expose it. It's the mirror image of /providers.

Both are available inside the unified OpenAI-Compatible API plugin and as standalone backend plugins you can put on their own route:

• cp:...aigateway.plugins.LlmProvidersCatalog
• cp:...aigateway.plugins.LlmModelCapabilities

# every provider that can do BOTH image and text
curl "https://api.domain.tld/v1/providers?capabilities=image,text" \
  -H "Authorization: Bearer $TOKEN"

GET /providers accepts a repeatable (or comma-separated) capabilities filter and returns only the providers that expose all of the requested capabilities.

// GET /providers
{ "object": "list", "data": [
  { "id": "openai", "label": "OpenAI", "capabilities": ["text","audio","image","embedding","moderation"] },
  { "id": "ovh-ai-endpoints", "label": "OVH AI Endpoints", "capabilities": ["text","audio","image","embedding","moderation"] }
]}

// GET /model-capabilities
{ "object": "list", "data": [
  { "id": "embedding", "label": "Embedding", "providers": ["openai","mistral","cohere","ovh-ai-endpoints","..."] }
]}

Broader provider coverage

New embedding providers

• OpenAI Compatible — bring your own OpenAI-compatible embedding endpoint, with a configurable display name (provider_name) and optional custom headers / param_mappings.
• OVH AI Endpoints 🇫🇷 🇪🇺 — via their unified OpenAI-compatible API.

OVH AI Endpoints 🇫🇷 🇪🇺 now also supports, through its unified OpenAI-compatible API:

• Image generation & editing
• Audio transcription (speech-to-text)
• Moderation

OpenAI Compatible is now available for image, audio and moderation too (in addition to chat and embeddings) — point it at any OpenAI-compatible endpoint, give it a name, and you're done.

Release Infos

• the documentation is available here
• release is available here

Contributors

• @mathieuancelin