Sim Atlas

Search and discovery platform for simulation nodes — Python functions, workflow definitions, and pyiron nodes used in scientific computing.

Repository layout

sim_atlas/
├── backend/    FastAPI server (sim-atlas-backend)
├── frontend/   Search & discovery SPA, served at /
├── web_ide/    Visual workflow composer SPA, served at /ide
├── toolkit/    Client-side parser & upload library (sim-atlas-toolkit)
└── docs/       Architecture docs and ADRs

How it works

1. Researchers run the toolkit on their machine to parse their Python packages and push node metadata to the server.
2. The backend stores metadata, generates AI embeddings and AI-refined docstrings on demand, and exposes a REST API and MCP tool.
3. The frontend and web_ide are React SPAs built into the backend's static directory and served by it.

Quickstart

1. Start the server via uvx

cd <sim-atlas>
uvx sim-atlas

On the first start the server will create a configuration file in the current working directory if it does not find one.

The API and both SPAs are now available at http://localhost:8000.

2. Generate an API token

Write access (uploading nodes) requires a JWT token:

cd <sim-atlas-env>
uvx --from sim-atlas sim-atlas-access-token "Your Name" "you@example.com"
export SIM_ATLAS_API_URL=http://localhost:8000/api/v1
export SIM_ATLAS_API_TOKEN=<token>

3. Upload nodes from a Python package

Load the environment with the modules you want to upload. Install the sim-atlas-toolkit into this environment. And upload the module. For example:

uv pip install sim-atlas-toolkit
uv run sim-atlas-upload --recursive filesystem mypackage.mymodule

4. Explore your newely created nodes

Head over to the webinterface and explore the nodes you uploaded:
http://localhost:8000