Otoroshi 17.15 brings agents with tools and memory, OAuth2 token exchange, PluginPresets and requires Java 25
Otoroshi v17.15.1 is available. It introduces an OAuth2 token exchange plugin implementing RFC 8693 for secure service-to-service API communication, a new PluginPreset kind to package and reuse plugin configurations, a MandatoryConsumerPreset plugin and a workflow-based authentication module. The OAuth2Caller plugin can now rely on an authentication module for the client credentials flow, and $2b$ / $2y$ Bcrypt hashes are now accepted.
A redesigned admin interface and manual
The admin interface has been redesigned with refreshed themes, a new home page, a reworked API editor with sticky actions and form views capped at 1000 px to improve readability. The UI also now only displays entities a user is actually authorized to see. The full manual has been reworked as well, with ditaa schemas replaced by mermaid ones, and the OpenAPI specification and Swagger UI can be protected behind an optional access secret.
API management, data exporters and plugin development
API management gains direct subscription to a plan and a new consistency service. The PostgreSQL data exporter now supports retention and plugin developers can rely on a reusable stateful clients internal API. Several fixes address a race condition in data exporters loading, legacy audit event layouts, body consumption handling and the OAuth2Caller password grant type.
LLM extension: agents, memory and Kreuzberg
This release includes LLM extension 0.0.75, which significantly expands agent capabilities with built-in tools (system_exec, file_read, file_write, http_call), a per-session working memory, autonomous persistent memory on Redis and PostgreSQL and a new agent OpenAI proxy plugin. Embedding stores and persistent memories now support many backends (PostgreSQL, Redis, Elasticsearch, OpenSearch, Qdrant, Weaviate, Pinecone, ChromaDB), a semantic cache on Redis with a custom embedding model and MCP audit events. The Anthropic provider exposes version and beta settings, and Kreuzberg is now available as a content-to-markdown and OCR backend.
Java 25 is now required
Otoroshi 17.15 requires Java 25 to load the Kreuzberg content-to-markdown and OCR backend. Starting with this release, all new Otoroshi add-ons are created with CC_JAVA_VERSION=25, and any update to Otoroshi 17.15 automatically switches the underlying Java application to Java 25. If you prefer to handle this manually, you can set CC_JAVA_VERSION to 25 on the Java application backing your Otoroshi add-on and rebuild it before updating Otoroshi itself.
You can update through your add-on dashboard in the Clever Cloud Console. You can also set CC_OTOROSHI_VERSION of the underlying Java application to v17.15.1_1776891995 and rebuild it, or use Clever Tools:
clever features enable operators
clever otoroshi version check yourOtoroshiNameOrId
clever otoroshi version update yourOtoroshiNameOrId
clever otoroshi version update yourOtoroshiNameOrId v17.15.1_1776891995