Skip to content

🤖 DV-Agentic-System

Core AI Agentic System for UVM (Universal Verification Methodology) & pyuvm Hardware Verification

Welcome to the official developer documentation for dv-agentic-system — a state-of-the-art, multi-agent artificial intelligence framework engineered to automate and optimize the design verification pipeline.

🚀 Key Capabilities

Our agentic ecosystem is designed to solve complex hardware verification challenges through collaborative specialized sub-agents:

  • 🧠 Multi-Agent Collaboration: Managed orchestrator coordinates spec analysts, code generators, sim controllers, and log analyzers.
  • ⚡ Intelligent Code Generation: Seamlessly generates target stimulus, checkers, and test sequences with UVM/pyuvm best practices.
  • 📈 Closed-Loop Debugging: Automatically executes simulator runs, parses logs, isolates root causes, and refines code until convergence is achieved.
  • 🔍 Semantic Specification Analysis: Parses design specifications to extract test intent and coverage goals.

🏛 Architecture & Design Tenets

The system is constructed with strict engineering rigor to ensure reliability, safety, and performance:

  1. Model-Tool-Instruction Separation: Clean abstraction layer decoupling AI logic, execution environments, and system instructions.
  2. Dynamic Escalation Guardrails: Prevents token budget burn by escalating immediately to human review when circular loops or shifting failure spaces are detected.

🗺 Documentation Guide

Discover our documentation through three main tracks:

graph TD
    Home[Docs Home] --> Guide[Architectural Guides]
    Home --> Reference[API Reference]

    Guide --> G1[System Overview]
    Guide --> G2[Structure & Design]
    Guide --> G3[Prompt System]

    Reference --> R1[Agents API]
    Reference --> R2[Tools & Adapters]
    Reference --> R3[CLI Command Docs]

📘 Architecture

  • System Overview: Read about the high-level workflows and multi-agent topology.
  • Structure & Design: Deep dive into classes, states, and the orchestrator decision lifecycle.
  • Prompt System: Learn how templates and system instructions are loaded dynamically based on task context.

⚙️ API Reference

  • Agents API: Reference details for orchestrator and sub-agents.
  • Tools & Adapters: Documentation on simulation adapters, LLM interfaces, and environment control.
  • CLI Reference: Commands to run tests, compile docs, and perform manual triage.

🏁 Quick Start

To install dependencies and build documentation locally:

# Clone the repository
git clone https://github.com/anlit75/dv-agentic-system.git
cd dv-agentic-system

# Install dependencies using uv
uv sync --all-groups

# Build and serve documentation locally with live-reload
uv run mkdocs serve