CYTools Interface

CYTools Interface#

This notebook explains the boundary between StringForge and CYTools. StringForge does not replace CYTools and does not try to duplicate toric-geometry computations. Instead, it stores catalogue rows and cached derived data in a form that can be handed back to CYTools, or forwarded to JAXVacua when the goal is flux-vacuum physics.

What belongs where?#

Task	Preferred tool
Query the public TDF catalogue, cache parquet shards, and run offline/HPC workflows.	StringForge
Reconstruct or inspect toric objects such as polytopes, triangulations, and cones.	CYTools
Build LCS periods, EFTs, and flux-vacuum finders from the stored data.	StringForge bridge into JAXVacua

Use CYTools directly when you want to perform new toric-geometry computations. Use StringForge when you want reproducible database access, cache management, and stable handoff to downstream physics packages.

1. Query a toric row#

The pure database layer is enough for catalogue queries and polytope recovery. No JAXVacua model is constructed in this step.

from stringforge import TDFDatabase, LCSDatabase

raw_db = TDFDatabase()
rows = raw_db.query(h11=2).head(3)
rows[[c for c in ("h11", "h12", "ks_id", "triang_id") if c in rows.columns]]

2. Recover polytope data where supported#

For TDF rows, get_polytope returns the stored polytope points and any additional polytope-level fields present in the parquet split. This is the StringForge side of the handoff: it retrieves data reproducibly and leaves any further toric computations to CYTools.

row = rows.iloc[0]
polytope_data = raw_db.get_polytope(
    ks_id=int(row["ks_id"]),
    h11=int(row["h11"]),
    h12=int(row["h12"]),
)
polytope_data.keys()

If CYTools is installed, the stored points can be passed to CYTools explicitly. The import is kept optional because StringForge’s database and vault workflows should remain usable without CYTools.

# Optional CYTools handoff:
# from cytools import Polytope
# polytope = Polytope(polytope_data["polytope_points"])
# polytope.points()

3. Hand the same row to JAXVacua#

When the next step is flux-vacuum physics rather than toric inspection, use LCSDatabase. This layer presents Hodge numbers in the mirror convention expected by JAXVacua and can construct either an lcs_tree or a FluxVacuaFinder.

lcs_db = LCSDatabase(dataset="tdf")

tree = lcs_db.load(
    ks_id=int(row["ks_id"]),
    triang_id=int(row["triang_id"]),
    # raw_db uses catalogue convention; LCSDatabase uses mirror convention.
    h11=int(row["h12"]),
    h12=int(row["h11"]),
    include_gv=False,
    include_conifolds=False,
)

type(tree).__name__, tree.h11, tree.h12