Skip to content

Jaxon Platform Documentation

Welcome to the Jaxon platform documentation. Jaxon verifies text content against formal policies using logic-based rules, delivering mathematically provable compliance results rather than probabilistic judgments.

Getting Started — A walkthrough of the platform's core workflow, from documents to rulesets to test results.

Concepts — Learn the core building blocks of the platform:

  • Documents — Text content that the platform analyzes
  • Rulesets — Rulesets that verify documents against policy requirements
  • Datasets — Collections of test data for measuring ruleset accuracy
  • Runs — Execute rulesets against documents and review results
  • DSAIL Language — The formal logic language behind the verification engine
  • The Corpus — A governed vocabulary of claims, the @namespace.name sigil a rule names one with, the axioms that connect rules across sources, and ordered verdicts
  • Analysis — Asking the corpus what follows, what would have to change, and what the rules can do at all
  • Vocabulary Workbench — The claim registry, the merge queue, promoting a rule's claim into the shared vocabulary, and authoring claims, corpus rules and axioms
  • Corpus Health — Static verification of the rules against each other

Tutorials — Hands-on tutorials for building real compliance verification setups.

Reference:

Security — Threat model, encryption, access control, and audit posture.

What is Jaxon?

Jaxon verifies LLM outputs and text content against formal policies. DSAIL evaluation compiles policy requirements into formal logic and evaluates them with an SMT (Satisfiability Modulo Theories) solver, producing mathematically provable results — a rule either holds, is violated, or the text lacks sufficient information to determine the outcome.

The platform is where you configure projects and build rulesets — translating your policy documents into formal verification logic. Once your rulesets are developed and tested, they can be deployed directly through the Jaxon API. The platform itself is your development and testing environment: upload sample documents, run rulesets against them, and iterate until your verification logic is ready for deployment. The platform handles the complexity of translating natural language policies into formal logic, extracting relevant information from text, and producing definitive compliance assessments.

The corpus is a design-time layer beside all of that: a governed vocabulary of claims plus the axioms that connect rules written from different sources, so the platform can reason about the whole body of rules together — finding contradictions between two policies, marking a document at the level the rules force, and answering what would have to change to reach a different outcome. Its axioms and verification never touch a run. Its vocabulary does, in exactly one way: a rule that names a claim from the registry as @namespace.name is asked that question once, and every rule naming it reads the same answer. See The Corpus for the concepts, One claim, one answer for what that means in a run, Corpus Health for verification of the rules against each other, and Analysis for asking questions about a project's runs.

Release Notes — See what's new in the latest version.