0.0.82

Otoroshi LLM Extension — v0.0.82

New

• Provider & capability catalogs on the Otoroshi admin API — the /providers and
  /model-capabilities catalogs are now also reachable through the Otoroshi admin API
  (authenticated with an admin API key), in addition to the unified OpenAI-compatible plugin:

  • GET /api/extensions/cloud-apim/extensions/ai-extension/providers
  • GET /api/extensions/cloud-apim/extensions/ai-extension/model-capabilities

  The capabilities filter and the response shape are identical to the plugin endpoints.

• Rampart body redaction endpoint — a new terminal backend plugin
  (RampartBodyRedactionBackend) that reads the request body, redacts PII with the local
  Rampart model, and returns the redacted body directly as the response (same
  Content-Type, X-Rampart-Redaction-Applied header). Turns any route into a
  self-contained redaction API. Shares the config and fail-open behavior of the existing
  Rampart request/response body redactors.

Improvements

• Mistral — chat messages are now sent using the OpenAI content representation,
  improving handling of structured / multi-part message content.

Release Infos

• the documentation is available here
• release is available here

Contributors

• @mathieuancelin