Reuse and share runs

Three days later you need the quench you ran on Tuesday — to check a number, to run the same cycle on a different grade, or to send it to the colleague who asked for it. Everything you ran in a session is still there, numbered, and reachable from the conversation itself.

How runs are numbered#

Every simulation that completes takes the next run version in the session: run 1, run 2, run 3. Versions are never reused. If you delete run 3, the next run is still 4. Artifacts for each run live in a run-scoped folder under a session-scoped prefix, with the run version zero-padded to three digits — run 3 stores its artifacts under run_003.

Where in the run the number gets reserved differs by simulator, and it shows up as gaps. A heat-treatment run reserves its version before it writes anything, so a run that fails partway still consumes a number. A CCT sweep reserves its version only once the diagram has been produced, so a sweep that fails before that point leaves no gap. Either way the reservation is atomic: two runs in the same session never collide on a number, and a retired number never comes back.

A CCT sweep of five cooling rates is one run, not five. A spot welding sweep of five currents is five runs, because each was a separate execution.

List what you have#

List my runs.

Phases replies with a count and, for each run, its version, its storage prefix, the simulator it was inferred to be, its status, its summary line, and a map of which output files are present. A failed run is listed like any other, and for a failure the summary line carries the reason — so "what went wrong with run 4" is answerable without re-running it.

Filter when a session is busy:

List my CCT runs.

List my RSW runs.

Under the hood the filter normalizes a handful of aliases before matching — ht, phases, phase and heat_treatment are the same simulator, as are vgleeble, v_gleeble and v-gleeble — so it does not matter which of those names the agent picks up from your wording. The filter is applied first and the most-recent cut second, so List my last 5 CCT runs gives you the five most recent CCT runs rather than the CCT runs among the last five of everything.

By default you get the most recent 20. Asking for "my failed runs" does not narrow anything — a status value is accepted and echoed back in the result, but it is not applied as a filter, so read the status column yourself.

Inspect one#

Inspect run 3 and tell me its status, summary and configuration.

inspect_run returns the run version, storage prefix, simulator type, status and summary, the metadata the run recorded, and a presence map of the run's artifacts. Asking without a number gives you the highest run version in the session. Inspection searches the most recent 100 runs of the session; in a session longer than that, an early version answers No matching runs found. even though the run still exists.

Two limits are worth knowing before you plan around it.

It does not return numeric file contents. It tells you which artifacts exist and hands back the metadata the run recorded — for a CCT sweep that metadata includes the rates and the final phase fractions per rate. It is not a way to read a results table row by row.

It only ever returns heat-treatment plot links. Inspection signs a URL only for an image the presence map below reports, and the map holds just five — Time_phase.png, Time_dilatation.png, Temp_dilatation.png, Time_statevars.png and Time_strain_stress.png. A CCT diagram and an RSW nugget animation are never on that map, so asking the agent to show you an earlier CCT or spot-weld run's figures comes back empty. Those images reach you when the run itself posts them into the conversation, and afterwards through the Runs panel or the Results tab described below.

The presence map is a fixed checklist of fourteen names, not a directory listing: PhaseLog.txt, Time_phase.txt, Time_dilatation.txt, Time_statevars.txt, Time_strain_stress.txt, mat_Steel.k, mat_Steel_top.k, mat_Steel_bottom.k, outputs.zip, and the five PNGs above. A false means that particular file is not at the top of the run folder. A file that is not on the list at all — a CSV, a VTK frame, anything in a subfolder — is simply not reported, so for a CCT or V-Gleeble run the map can read as all-false while the run produced plenty. Use the browsing surfaces to see what is actually there.

Refer back in a later turn#

The active session material persists in a session, so a follow-up needs no re-loading:

Run the same heat treatment on 22MnB5 instead.

Two things stay true when you do this. Phases runs only the simulator you named in that message, and it will not re-run a simulation merely because you asked about an earlier one — inspecting run 3 does not authorize a new run. If you want a new run, say so.

For spot welding there is a specific reuse path: when microstructure tracking is enabled, a run can reuse a previous run's plate material card per plate. Outside that case, material reuse works through the session material or a saved workspace material, not through the run.

Browse the artifacts#

Signed in, you have two browsing surfaces: the Simulation Runs panel behind the Runs button, which lists the current session's runs and their files, and the simulation panel's Results tab, which lists them as rows with View / Hide, Reconfigure and delete actions. Both are described in full in Sessions, runs and artifacts.

Both list the run's complete file set — there is no image filter on an ordinary account. A run's owner sees PhaseLog.txt, material_resolution.json, the mat_*.k deck and outputs.zip alongside the plots.

Both also put a download button on every file row, so getting one file out is one click: a plot, a numeric text file, an LS-DYNA deck or the whole outputs archive.

In the Runs panel you can read a file without downloading it. Click any row and it opens in an inline viewer below the list: images render, and the text outputs — .txt, .json, .csv, .k, .log, .yaml, .yml, .md — open as text in a scrollable block with its own download button. That is the route to reading a results column yourself rather than asking the agent to summarise it. The Results tab does not have the inline text viewer; expanding a row there shows the recorded configuration, the image gallery and the file list.

Guests do not get any of this. The Runs panel, the Materials pages, the simulation panel and the chat-history sidebar all require a signed-in account — see Guests and limits.

Sharing in Phases is per conversation, not per run. In the left sidebar, open the … menu on the chat you want and choose Share. A dialog titled Share link to chat appears, with the description "Anyone with the URL will be able to view the shared chat.", the chat title, its date, the /share/ path as a clickable badge once the link exists, and a Copy link button. Copying puts the URL on your clipboard and confirms with a "Share link copied to clipboard" toast.

The link is a /share/ URL ending in the chat's own id.

What the recipient sees on that page: the chat title as a heading, a line giving the creation date and the message count, and the full rendered conversation. What they do not see:

No sign-in is required, so treat the URL as public to anyone who has it.
The configuration cards still render on a shared page, but they are inert — the fields are visible and pressing Run Simulation does nothing. Your reader gets the conversation and the result images, not runnable controls.
The Runs panel and artifact browsing are not on the shared page.
There is no un-share control in the product interface. Share deliberately.

If the conversation should not be public, send the results another way rather than the link.

Delete#

The same … menu has Delete, guarded by an "Are you absolutely sure?" confirmation warning that this permanently deletes the chat and removes the data. Clear chat history in the user menu does the same across your history and can report a partial result when some sessions still hold active simulation data.

Deletion is refused while a run is still executing or an artifact write is in flight; retry once the run reaches a terminal state. Deleted session IDs and run versions are permanently retired.

One caveat worth knowing about links to individual files. Artifact links are signed and time-limited, and the window depends on where the link came from: a plot link the agent posts into the conversation is signed for seven days, while a view or download link minted by the Runs panel or the Results tab is signed for one hour. Either way, deleting the run deletes the stored files themselves, and an outstanding link then points at nothing and stops resolving. Until it expires or you delete the run, treat an issued artifact link as live.

Variations#

Pick up a comparison. List my runs, then Compare run 2 with run 5. Asking to be shown two earlier CCT diagrams does not work — inspection returns no CCT images — so open them in the Runs panel side by side, or let Phases do the arithmetic. See Compare and analyse runs and Compare two grades.

Re-run with one change. Use Reconfigure in the Results tab to load an old configuration into the Configure tab, adjust one field, and run again. This is the fastest way to extend a weld schedule sweep.

Keep the material, not just the run. If a chemistry is worth reusing across sessions, save it to your workspace rather than hunting for the run that used it.