embodied-ai-architect 1.5.0

Hardware/Software Codesign Space Exploration and Optimization Environment for Embodied AI systems

Embodied AI Architect

A design environment for creating and evaluating autonomous agents, with hardware/software codesign space exploration and optimization.

Features

• Model Analysis: Analyze PyTorch model structure and compute requirements
• Hardware Profiling: Recommendations for edge/cloud deployment
• Multi-Hardware Benchmarking: Local CPU, remote SSH, Kubernetes backends
• Interactive Chat: Claude-powered architect for design decisions
• Codebase Analysis: Scan and assess application codebases for hardware deployment
• SoC Optimization: LangGraph-based RTL optimization loop (experimental)

Installation

pip install embodied-ai-architect

With optional dependencies:

# Remote SSH benchmarking
pip install embodied-ai-architect[remote]

# Kubernetes benchmarking
pip install embodied-ai-architect[kubernetes]

# Interactive chat (requires ANTHROPIC_API_KEY)
pip install embodied-ai-architect[chat]

# All optional dependencies
pip install embodied-ai-architect[all]

Usage

# Show available commands
branes --help

# Analyze a PyTorch model
branes analyze model.pt

# Run full workflow
branes workflow run model.pt

# Benchmark on local CPU
branes benchmark model.pt --backend local

# Scan and assess a codebase for hardware deployment
branes codebase scan /path/to/project
branes codebase assess /path/to/project --hardware jetson_orin

# Interactive chat session
export ANTHROPIC_API_KEY=your-key-here
branes chat

Environment Variables

Variable	Description
ANTHROPIC_API_KEY	Required for Claude-powered features (chat, codebase analysis)

Documentation

For full documentation, development setup, and contributing guidelines, visit the GitHub repository.

Related Projects

• embodied-schemas: Shared Pydantic schemas and hardware catalog
• graphs: Analysis tools and roofline models

License

MIT License