Skip to content
phynite-lab
Phynite DocumentationDownload

Run history, research context, and evidence

Use Run history for execution attempts and cancellation. Use Research context for the project timeline, compact run ledger, recent comparison, and read-only AI context. These panels report retained evidence; they do not replace solver logs or scientific validation.

Complete guide

Find the record for the current project

Open full resolution

Inspect the selected project and first recorded cause; no solver output is invented.

Inspect the selected project and first recorded cause; no solver output is invented.
  1. Open a project’s Canvas. Choose the clock/history icon labeled Run history to see runs, their creation times, per-module attempt states, and recorded errors.
  2. Inspect a specific module’s Execution result for numerical summaries, Execution log, artifact downloads, and trajectory preview when available. An old execution can be marked Outdated after its current setup changes.
  3. Use Cancel branch for an active module branch or Cancel run for an active workflow. Wait for the terminal state before resubmitting; a cancellation request is not yet a completed stop.
  4. Choose the activity-line icon labeled Research context. Refresh research context reloads recorded evidence for this project.

Read the ledger, timeline, and latest summary

Run ledger shows up to the latest eight records with their names or kinds, timestamps, status, recorded failure category/reason, and input checksum where available. Timeline shows up to the latest eight recorded project events. Empty lists mean no evidence was returned, not that an unrecorded calculation succeeded.

Summarize latest run displays the latest record’s status and available result summary, checksum, allocated CPU/RAM, and whether provenance was recorded. Missing values remain unavailable. The compact summary is assembled from saved records; it is not a new AI calculation.

Allocated CPU and memory describe the job’s requested or effective budget. They are not measured CPU utilization or peak RAM. Use the actual execution record and logs for recorded timings and resource failures.

Compare recent runs without overinterpreting them

  1. Complete at least two runs in the project. Compare recent runs remains disabled until two completed records are available.
  2. Choose Compare recent runs. The current compact comparison selects the two most recent completed runs and displays status, input checksum, and completion time.
  3. Open the original input and result artifacts before comparing energies, convergence, trajectories, or runtimes. The panel does not automatically prove equal structures, force fields, pseudopotentials, normalization, or physical conditions.
  4. For a scientific overlay or parameter series, prepare compatible numerical datasets in Figure Lab and record units and reference conventions. Use separate panels when the sources are not comparable.

Give local AI traceable research context

Local AI means the connection and credential handling run on your computer. The selected provider can process the submitted prompt and context under its own service terms. Keep confidential research out of a provider request unless you intend to share it with that provider.

  1. Connect an available local provider in Settings → AI provider accounts. Cloud login is not required for this connection.
  2. In Research context, choose Add context to local AI. The project context appears as a read-only source in the AI workspace.
  3. Ask a bounded question, such as identifying the latest failure or explaining a recorded input difference. Ask the response to cite the run or artifact it uses, then inspect that evidence yourself.
  4. Review any proposed scientific change through the existing approval flow. Adding read-only evidence does not grant permission to edit a structure, execute a solver, or change a potential.

Failure evidence, notifications, and team visibility

Enable local job/runtime alerts in Settings → Notifications to be informed of actual state changes while the app is open. An alert is a pointer to the saved record; keep the log and provenance before retrying.

SignalInterpretation
Preflight or setup rejectionFix runtime, isolation, source, potential, or validation prerequisites before a new run. No solver log exists if the solver never started.
Solver, resource, or artifact failureRead the recorded first cause and relevant log. A failure while saving artifacts does not imply the numerical process never ran.
Recover saved result appearsThe local app found recoverable output. The action verifies and imports saved results; it does not rerun the solver.
Local context onlyNo project sync is enabled. Signing in to Cloud alone does not publish this project.
Team visibility enabledThe project already has approved Cloud Sync scope and a reported revision. The panel does not create sharing permissions or upload additional files.

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