anormang1992/vre

<img width="2048" height="2048" alt="vre_logo" src="https://github.com/user-attachments/assets/9419c6a6-4be6-418f-8199-090bdf5437a9" />

VRE — Volute Reasoning Engine

Epistemic enforcement for autonomous agents.

VRE is a Python library that gives autonomous agents an explicit, inspectable model of what they know before they act. It is not a permissions system, a rules engine, or a safety classifier. It is a mechanism for making an agent's knowledge boundary a first-class object — one that can be queried, audited, and enforced at runtime.

---

Table of Contents

• The Problem
• How It Works
  • The Epistemic Graph
  • Relata
  • Knowledge Gaps
  • Layered Safety
• Scope
• Getting Started
  • Installation
  • Infrastructure
  • Seeding the Graph
• Core API
  • Connecting to VRE
  • Agent Identity
  • Checking Grounding
  • Using the Trace as Agent Context
  • Checking Policy
• The vre_guard Decorator
  • Parameters
  • Execution Flow
• Callbacks
  • on_trace
  • on_policy
• Learning
  • How It Works
  • Candidate Types
  • Provenance
  • Reachability Prerequisites
  • Reference Loop
• Policy System
  • Defining Policies
  • Policy Callbacks
  • Evaluation Flow
  • Policy Wizard
• Integrations
  • [LangChain + Ollama Reference Agent](#langchain--