AtomSculptor

Materials-science agent workspace using Google ADK + Anthropic Sandbox Runtime (srt).

Requirements

• Linux or WSL2
• Python 3.12+
• Node.js 18+ / npm
• On Debian/Ubuntu, ./install.sh will automatically install bubblewrap, socat, ripgrep, and the Anthropic sandbox runtime CLI.

---

Quick Install

git clone <repo-url> && cd AtomSculptor
./install.sh

This creates a .venv, installs Python packages, installs the sandbox runtime CLI (srt) and its Linux dependencies when supported, installs frontend dependencies, and sets up .env.

Flags:

Flag	Effect
--all	Install all extras (web + dev + treesitter-full)
--dev	Include development tools (ruff, mypy, pytest, …)
--no-venv	Skip venv creation; use current environment

---

Manual Install

sudo apt-get install -y bubblewrap socat ripgrep

python3 -m venv .venv && source .venv/bin/activate

# Core + web GUI (recommended)
pip install -e ".[web]"

# Optional extras (combine as needed)
pip install -e ".[dev]"               # ruff, mypy, pytest
pip install -e ".[treesitter-full]"   # JS/TS/Rust/Go/Scala/Java/C++ parsers

# Frontend
cd web_gui/static && npm ci && cd -

# Sandbox runtime CLI
npm install -g @anthropic-ai/sandbox-runtime

# Environment
cp .env.example .env   # then edit with your API keys

---

Configuration

Edit config.yaml to set model backends and sandbox paths:

PLANNER_MODEL: "openai/qwen3-max"
SANDBOX_DIR: "sandbox/.runtime"

Set credentials in .env:

OPENAI_API_KEY="<your_key>"
OPENAI_API_BASE="<your_base>"
MP_API_KEY="<your_mp_api>"

---

Run

source .venv/bin/activate

# Web GUI (recommended)
python main.py --web

# ADK CLI
python main.py
# or equivalently:
adk run agent_team

---

Code-Graph-RAG Integration

Optional code analysis powered by Memgraph. Requires Docker.

# Start Memgraph
docker compose up -d

# Or manually:
docker run -d --name memgraph -p 7687:7687 memgraph/memgraph-platform

Set TARGET_REPO_PATH in config.yaml to point at the codebase to analyze (e.g. pymatgen, ase, rdkit sources). Install full tree-sitter support for multi-language parsing:

pip install -e ".[treesitter-full]"