Roadmap

What we're working on and exploring next. Nothing on this page is a commitment, and there are no dates. Items move to the changelog when they ship.

Have a use case or a priority? Tell us in GitHub Discussions.

#Exploring

#kerniq-ai

A standalone library for building AI-powered Rust applications through one consistent API, without learning the details of every model provider, vector database or protocol. It would cover:

• Model APIs: chat and completion across providers behind one interface.
• Chat client: conversations, streaming and structured output.
• Embeddings and vector stores: storing and searching embeddings for retrieval-augmented generation (RAG).
• Tool calling: expose Rust functions to models as tools.
• MCP: connect to and serve tools over the Model Context Protocol.

kerniq-ai is planned as a separate project, not part of Kerniq Server. It should work with any Rust web framework, including Kerniq, Axum, Actix Web, Rocket and Poem, or with no web framework at all. Kerniq will provide a first-class integration.

It is at the design stage. The most useful input right now:

• which model providers and vector stores you need first;
• whether you'd use MCP mainly as a client, a server, or both;
• what you use today (and what you'd want done differently).

Share it in the kerniq-ai discussion.