Tutorials

This page collects the executable notebooks in one place. Use it as the main entry point once you want to run code rather than read background material.

Choosing a path

First orientation

Query a catalogue, load a model as lcs_tree, construct a JAXVacua FluxVacuaFinder, and learn where StringForge stops.

StringForge Quickstart

Database access

Query HuggingFace-hosted catalogues, load Calabi-Yau data, batch models, and manage cache/offline workflows.

Database Interface

Vacua storage

Validate, designate, store, and prepare vacuum datasets for the shared vacua_vault.

Vault workflow — produce, validate, designate, push

Cluster workflows

Older chunk-and-merge pattern; prefer Vulcan for new cluster runs.

Cluster-Parallel Flux Enumeration and Sampling

Production cluster runs

Stage vacuum batches on workers, batch-commit on a head node within HuggingFace’s commit-rate cap, query and stream the resulting dataset.

Vulcan: production vacuum-forging for cluster workloads

Tutorial catalogue

Quickstart and overview

Notebook	Use it for
Quickstart	The shortest StringForge-first database-to-JAXVacua workflow.
StringForge ecosystem pipeline	Public package choreography plus schematic planning notes for future packages.

Geometry input and databases

Notebook	Use it for
CYTools interface	Understanding the boundary between CYTools objects, database rows, and JAXVacua model data.
Complete Intersection Calabi-Yau Threefolds	Working with CICY catalogue data, with provenance caveats.
Database interface	Querying catalogues, loading individual models, batch loading, offline mode, and cache management.

Vacua vault and infrastructure

Notebook	Use it for
Vault workflow	Validating, designating, and preparing vacuum datasets for upload.
Storing vacuum objects	Writing jaxvacua.vacuum Vacuum / PFV / AFV objects directly (curated and Vulcan tiers) with exact, finder-free read-back.
Vacua storage	Local vacuum storage, querying, designation, retraction, and sharing.
Cluster parallelisation	Exporting scan chunks, processing them on a cluster, and merging results.
Vulcan cluster runs	Production vacuum forging: stage on workers, batch-commit on the head node, query and stream as an ML dataset.

Advanced curated subsets

Page	Use it for
KKLT Database	Specialised conifold-class indexing and tags for KKLT-style searches.