Project overview

Praxis AI

An alpha AI gateway with provider-aware routing and filters, built on the Praxis proxy runtime.

Praxis AI adds AI request classification, provider routing, translation, and agent-protocol features to the Praxis proxy runtime. Its releases are independent of the core proxy. The selected v0.4.1 release is alpha; its security policy supports the 0.4.x line as alpha.

Start with the container quickstart to route a request to OpenAI. It requires an OPENAI_API_KEY, a provider account, and network access to api.openai.com:443. For source builds and development, see the source quickstart (Praxis AI v0.5.0 (alpha)). For capabilities and setup-dependent features, read the release feature overview (Praxis AI v0.5.0 (alpha)).

The AI release and Praxis release numbers are independent. AI v0.4.1’s manifest depends on Praxis crates v0.7.0, while this site’s core default is v0.7.2; the site has not verified that pair as a compatibility set. Choose versions from the relevant project manifests and test the combination you deploy.

Default documentation: v0.5.0 (alpha)

Run Praxis AI

Praxis AI is alpha software; APIs and configuration may change before v1.0.0.

An inference request is classified, processed by AI filters, and routed to a model provider.
Praxis AI classifies and enriches inference requests before provider routing. Read the inference architecture. Read the explanation for Praxis AI

Choose what you need

How-to guides

Tutorial

Reference

Explanation