OpenGradient-SDK: Instructions file for Claude Code

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Project Overview

OpenGradient Python SDK - A decentralized model management and inference platform SDK. The SDK enables programmatic access to model repositories and decentralized AI infrastructure, including end-to-end verified AI execution. Uses uv for dependency management.

Development Commands

Build & Installation

# Install dependencies (syncs from pyproject.toml/uv.lock)
make install  # or: uv sync --all-extras

# Build distribution
make build

# Publish to PyPI (requires credentials)
make publish

Testing

# Run all tests
make test

# Run unit tests only
make utils_test

# Run integration tests
make integrationtest

Code Quality

# Run formatting (ruff, line-length=140) and type checking (mypy)
make check

# Generate documentation
make docs

CLI Testing

# Test model inference
make infer

# Test LLM completion
make completion

# Test chat functionality
make chat

# Test streaming chat
make chat-stream

# Test tool calling
make chat-tool

Architecture Overview

Core Components

1. Client (src/opengradient/client/): Client package with submodules

   • llm.py — LLM chat/completion with TEE and x402 payment support
   • model_hub.py — Model repository management (CRUD, upload)
   • twins.py — Digital twins chat integration (twin.fun)
   • alpha.py — Alpha Testnet on-chain inference and workflows
   • opg_token.py — OPG token Permit2 approval
   • tee_connection.py — TEE connection management (static and registry-based)
   • tee_registry.py — TEE endpoint discovery via on-chain registry
   • _conversions.py — Model input/output type conversions
   • _utils.py — Shared utilities (ABI loading, retry logic)

2. CLI (src/opengradient/cli.py): Command-line interface using Click

   • Commands: config, infer, completion, chat
   • Supports file-based input for messages and tools