Prism

explainable model routing engine

Prism

Decision engine for model routing.

Most LLM routers answer: “Which model?”

Prism answers: “Why this model?”

Read the ArchitectureGitHub

Architecture

A high-level overview of Prism's routing architecture

Prism architecture

Philosophy

Three ideas the engine is built around

Explainable

Every decision Prism makes is accompanied by the reasoning behind it. A model choice without a reason is a guess.

Provider Agnostic

Prism has no default vendor. Model profiles are data, not hardcoded assumptions, so new providers are additive, not architectural changes.

Architecture First

Interfaces and contracts are designed before implementation. The engine is built to be read, not just run.

Roadmap

Where the project is headed

Prism is early. This is the current sequence of work, not a set of dates.

  1. Foundation

    In progress

  2. Core Engine

    Planned

  3. Plugins

    Planned

  4. SDKs

    Planned

  5. Benchmarks

    Planned

  6. Public Beta

    Planned