Skip to content
phynite-lab
Phynite DocumentationDownload

Local troubleshooting and recovery

Use this short recovery path when the local launcher opens but a runtime, job, browser, or saved workspace does not behave as expected. Preserve the data directory and logs until the cause is clear.

Complete guide

Startup and browser

Diagnose preflight, solver execution, and result persistence separately
Concept diagram: investigate the first failing stage before retrying.
SymptomRecovery
A local service stopped during startupRun status and doctor with the same --data-dir. Read the startup logs printed or stored in the data directory; do not create a second workspace by changing paths.
The browser did not openUse --no-open and open the exact loopback URL printed by the launcher. Port 3000 may have been replaced with an available port.
A second command shows the existing instanceReuse the reported URL or stop the instance cleanly before starting a different data directory.
Projects appear emptyCheck the original --data-dir, status, and workspace identity. Reinstalling npm does not restore data from another directory.
Example / commands
phynite-lab status --data-dir ~/.phynite-lab
phynite-lab doctor --data-dir ~/.phynite-lab
phynite-lab --no-open --port 3000 --data-dir ~/.phynite-lab

Runtime and isolation

  1. Open Settings → Compute and verify that the requested DFT or MD/analysis runtime is installed, started, and capability-checked for this workspace.
  2. If a sandbox probe is unavailable, follow the operating-system remediation shown by the app. Do not disable user namespaces, bypass bubblewrap, or switch to an unverified runtime.
  3. For DFT, verify the selected pseudopotential library or exact UPF assets and their checksums. For MD, verify mapped elements, potential integrity, runtime validation, and the selected unit system.
  4. Run a small reference job once the Compute plan is ready. Keep the failed attempt and its first cause if it still fails.

Read the correct failure log

First failing stageEvidenceNext step
Preflight / setupVisible Compute plan or run-review reason; no solver log because the process never started.Fix runtime, isolation, source, potential, resource, or input validation, then submit a new attempt.
Solver executionExecution log, log tail, exit status, and retained input/artifacts.Read the first meaningful solver error; inspect memory, disk, syntax, and physical settings.
Result persistence / artifactSolver output may exist while result import or artifact verification fails.Keep the attempt, check disk and checksums, and use Recover saved result only when offered.
AI providerProvider settings status, capability probe, and safe connection error.Fix the official CLI login or provider availability. There is no silent provider fallback.

Release identity, license, and update verification

The documentation release metadata currently resolves phynite-lab@0.1.5. The first command reports your installed launcher; the second reads npm registry metadata. A newer registry version does not update an already running process.

Open Source code and license in the app for its corresponding source. A published distribution records its exact public source commit and source archive; release assets provide package checksums and qualification evidence. Phynite-owned local code is AGPL-3.0-only; third-party solvers and potential files retain their own licenses.

Finish or cancel active work, stop the app, update with npm install -g phynite-lab@latest, and reopen the same data directory. Verify the launcher version and Compute capabilities. Background update notices do not install packages automatically and a registry outage does not prevent local work.

Example / commands
phynite-lab --version
npm view phynite-lab version

Restart, cancel, and recover

  1. Cancel an active run from Run history and wait until it becomes Cancelled or another terminal state. Do not submit a replacement while the original is still active.
  2. After a launcher or computer restart, inspect the run state and saved log. Interrupted work requires review; the app does not silently rerun a scientific job.
  3. Use Recover saved result only when the app reports verified saved output. Recovery imports that output; it does not claim the solver converged or rerun the job.
  4. Before disk cleanup or migration, export the project archive and make an independent backup of inputs, provenance, checksums, and required QE tmp/wavefunction state.

Related topics

Your privacy

Cookie preferences

Choose whether Phynite may use optional analytics in this browser. You can change this anytime.

Waiting for your choice

This choice applies to this browser, including your workspace.

Read the privacy notice