Welcome to CHIA’s documentation!

CHIA is a framework for fast and flexible orchestration of hardware design workflows — where humans, agents, and existing hardware-design tools all work together.

New to CHIA? Start with the CHIA Basics page, or go straight to Installation to set things up.

The current release of CHIA is considered a beta. Many library nodes have not received thorough testing. Bugs throughout CHIA are likely.

(Co-designing Hardware and software with Intelligent Agents)

(Constructing Hardware with Intelligent Agents)

(CHipyard’s Intelligent Agents)

(CHoosIng an Acronym)

Getting Started

• CHIA Basics
• Installation
  • Optional extras
  • Core dependencies
  • Next steps
• Quickstart: Say Hello (World) with CHIA
  • What we will do in this tutorial
  • Requirements
  • 0. Project Setup
  • 1. print() ing Hello from a worker
  • 2. Print Hello in Verilator
  • 3. Asking an LLM to say Hello
  • 4. Editing RTL to say Hello

Concepts

• Architecture Overview
  • CHIA flows
  • CHIA clusters
  • Fault tolerance & reproducibility
  • Infrastructure

User Guides

• ChiaFunction
  • A few Ray concepts
  • Executing a CHIA node
  • Collecting results
  • Worker-side setup and cleanup hooks
  • Defining your own node
  • Wrapping actors (chia_actor)
  • Cross-cutting features
  • See also
• ChiaTool
  • A few MCP concepts
  • Anatomy of a tool
  • The tool lifecycle
  • Bridging tools back to nodes
  • Defining your own tool
  • Tool placement and resources
  • Advanced
  • See also
• Cluster Configuration Reference
  • Top-level structure
  • provider
  • auth
  • available_node_types
  • Cloud nodes
  • A mixed on-prem + cloud example
  • Command execution order
• Building a CHIA-compatible Docker image
  • What “CHIA-compatible” means
  • Extend a CHIA base image with your tool
  • Extend your tool image with CHIA
  • Wiring the image into a cluster
  • Pitfalls
  • See also
• Caching and Bypass
  • Tags
  • Bypass
  • Cache
  • Putting it together: populate, then replay
• Profiling
  • Enabling profiling
  • What gets recorded
  • Attaching custom metadata
  • Visualizing a profile
• Example: Agentic RoCC Accelerator (MemCpy)
  • What it does
  • Components
  • The accelerator contract
  • Running it
  • Tunable parameters
  • Debug feedback
  • Output and per-iteration diffs

CLI Reference

• CLI Reference
  • Cluster management
  • Visualization
  • Jobs
  • FireSim

API Reference

• Base Runtime
  • Bypass
  • Cache
  • Chia Function
  • Chia Wait
  • Chia KV Store
  • Colocated
  • Dispatch Proxy
  • PID Registry
  • LLM Call
  • MCP tool servers
• Chipyard Nodes
  • State definitions
  • Chisel build
  • CIRCT
  • Macro compiler
  • Chipyard / Hammer
  • RISC-V build
  • RISC-V objdump
  • RISC-V DV generation
  • Torture
  • Verilator run
  • Spike cosim
• Model Backends
  • Claude
  • Bedrock
  • Vertex
  • Antigravity
  • Codex
  • Opencode
  • Openai Compat
  • Openai Providers
  • Ollama
  • vLLM
• FireSim
  • Config
  • Chia Functions
  • State Def
• VLSI
  • Hammer
  • SRAM characterization (CACTI)
• Databases
  • Base
  • Postgres Node
  • Sqlite Node
• GitHub
  • Github Client
  • Github Issues Node
  • Github Pulls Node
  • State Def
• AWS
  • Config
  • EC2
  • Host
  • S3
• Simulators
  • gem5
• Analysis
  • Focal
  • PFAS
• LLM servers, providers, and agent Nodes
  • Base class
  • Providers

Indices and tables

• Index

• Search Page