6. Updating & troubleshooting

6.1. Updating LatticeMind

When you receive a newer wheel, open a terminal in the directory containing the wheel and upgrade in place:

python -m pip install --upgrade --force-reinstall <matching-wheel>.whl

The wheel installs the post-processing libraries (h5py, scipy, matplotlib) by default. Your projects in latticemind_outputs/ (or LATTICEMIND_ARTIFACTS_DIR) are untouched by an upgrade.

6.2. Troubleshooting

Symptom	Fix
“Provider configuration required” / missing API key	Set the key for your selected provider (see Choosing a provider) in `.env` in the folder you launch from, or as an environment variable.
Web UI port already in use	LatticeMind picks another free port automatically, or set `LATTICEMIND_WEB_PORT=7866`.
Web UI loads but CSS / live updates don’t (behind a proxy)	Set `LATTICEMIND_WEB_BASE_PATH` to your proxy prefix, e.g. `/proxy/7865`.
No browser / running on a server	Use the CLI: `LATTICEMIND_UI_MODE=cli latticemind`.
LatticeMind reports “License check failed”	Put exactly one valid RESCU `.lic` or `.license` file under `~/.nanoacademic/RESCU`, or set `LATTICEMIND_LICENSE_FILE` to the exact file. This startup check applies to every solver selection.
RESCU command needs configuration	Test `rescu` directly, or set `LATTICEMIND_RESCU_CMD` to the exact executable. The RESCU license configured for LatticeMind is also used for RESCU execution.
NanoDCAL command needs configuration	Test NanoDCAL directly, then set `LATTICEMIND_NANODCAL_CMD` or the Web UI’s Local NanoDCAL command field. NanoDCAL needs its own active solver license on the computer that executes it.
NanoDCAL neutral-atom data path needs configuration	Extract the NeutralAtomDatabase from the Nanoacademic Portal and set `LATTICEMIND_NANODCAL_DATA_PATH` or the Web UI’s NanoDCAL neutral atom data field to its root directory, not to the archive file.
A DFPT/phonon post-processing step fails in parallel	Some RESCU post-analysis stages are serial-only; LatticeMind strips MPI launchers for them automatically. See Settings & configuration reference.

6.3. Before launching expensive jobs

Use these guards to avoid surprises:

/cost              # estimate resource risk before launch
/launch-preview    # preview commands, inputs, readiness, serial guards, cost

RESCU and NanoDCAL feasibility depends on atom count, basis, k/q grids, functional, spin, wall time, memory, and scheduler limits. Start with a cheap validation run before launching dense DFPT, phonon, optics, Raman, relaxation, transport, or large-cell workflows.

6.4. Getting help

See Contact Us for the Nanoacademic forum and support email, and what to include in a report.