ASL PC Steward help center
Support: support@autosecurelogin.com. Tell us the app and Windows versions, the exact warning, when it first and most recently occurred (with timezone), what you expected, and your last successful action. Review any export before sharing. Never email passwords, fleet keys, private journals or unrelated files.
1. What Steward does by itself
PC Steward is a native Windows diagnostic and guarded-action application. It inspects performance, Windows events, startup services, scheduled tasks, sessions, sharing, storage and available cooling telemetry. It explains specific findings and available timing. A scan is not a repair, and a warning is not permission to terminate a process, format a drive or remove software automatically.
Local diagnostics do not require AI Hub, rBroker, Master Suite or a cloud account. The optional fleet connection reads an authorized coordinator. It does not make Steward the owner of remote GPU leases or CPU jobs. The combined Master Suite embeds the shared health workspace, while standalone Steward remains useful and keeps its own settings, history and protected-action records.
Use the Windows version declared by your installed package. Steward targets 64-bit Windows 11 build 22621 or later; a different ASL product's Windows 10 compatibility does not make this package compatible with Windows 10.
2. Run a scan and understand completeness
Start a scan from Overview. Ordinary read-only collection does not require administrator approval and does not install a driver or change Windows settings. Read both the findings and collection notes. A complete scan means the requested sources finished under their stated support rules, not that every possible hardware sensor exists or that every issue on the computer has been ruled out.
The health score summarizes current checks from that scan. Retained crash/update history, recovered events, and stale or unconfirmed evidence remain visible but subtract no points from the current score. Many old incidents cannot turn a healthy current sample into an immediate emergency. This does not prove those incidents were repaired: review their exact dates and recheck the affected app. Current resource pressure and current actionable system findings still affect the score. It is not a performance guarantee or a certification of the machine.
Partial means at least one source did not return usable information. Expand the note to see the source and reason: missing optional software, denied access, unsupported hardware, timeout or a provider failure are different problems. Missing data is not zero and is not a clean bill of health. A slow Windows inventory provider can recover later; an unsupported sensor cannot be enabled just by repeatedly refreshing.
Read sample time, scan age and inventory time separately. An old snapshot does not show what is happening now. Refresh deliberately after a relevant change. Do not disable security software or reboot merely because one optional source is unavailable; first identify the exact failing source and evidence.
3. Active Monitor is optional
Active Monitor starts off. Enable it from the appropriate screen when you need repeated readings, then turn it off when finished. The selected interval is a target: slow collection can take longer. Steward does not overlap scans simply to catch up. Inventory is not recollected at the same frequency as live counters.
Overview, fleet monitoring and temperature monitoring have separate scopes. Leaving a live page stops its page-owned monitoring as described there. An open status window is not required for a separately installed broker service to run. These ordinary application refreshes are not AI conversations consuming tokens.
4. Expand an issue before acting
Click a finding to read its explanation, exact affected target, observed values, available event timing and suggested next step. Last checked, last ran, last failed and most recent occurrence are not interchangeable. If Windows did not provide the event time, Steward must say unknown rather than invent it.
A historical crash or service failure may explain an earlier slowdown without proving it is still happening. Compare the event timestamp with current health. A high memory allocation can be intentional work. A scheduled task's last result does not identify its whole process tree or prove that another application caused a crash. Correlation is useful evidence, not automatic causation.
Prefer the recommended low-impact check first. Inspect the named service, event, process or task and its owner. Review unsaved work before any action that could end a user session. Avoid broad cleanup or changing unrelated services to silence a warning whose cause has not been established.
5. System RAM, GPU RAM and unfamiliar drives
System RAM is the computer's main memory. GPU RAM is dedicated graphics memory. They have different capacities, consumers and admission rules. A leased GPU commitment is not necessarily the driver's current physical allocation. Windows GPU allocation references are not all exclusive dedicated VRAM; they must not be added together as an extra reserve.
The broker applies its configured allowance once. OMEN uses one 5,120 MiB Windows allowance; customers should review their own hardware and policy rather than assume the same usable capacity. Steward displays evidence and timing; it does not recalculate a second allocation budget or hand out leases.
On Storage, inspect type, provider, size and readiness before identifying a drive. A cloud quota, removable volume, network share and physical partition are different things. A low-space warning is not permission to format or move data. Steward does not format disks or automatically choose projects/models to delete.
6. Connect to rBroker and AI Hub work
On Broker & Fleet, enter the trusted HTTPS coordinator origin and an authorized viewer key. The hostname is configurable. Connect deliberately and review the reported computers. A view is not enrollment or execution permission. Do not paste a node's private key or provider token into the viewer field.
Use the fleet page's own monitor for repeated updates. Leaving the connection page clears its in-memory key as described in the interface. Read fresh, stale, snapshot-only and not reported literally: an older coordinator may not provide expiry, and an expired report does not prove that its jobs stopped.
New compatible reports show the actual task title, project, purpose, stage and expected result, alongside IDs for correlation. System-RAM commitments and CPU worker state are separate from GPU leases. Older records without descriptions remain explicitly incomplete; Steward does not invent private task details.
rBroker owns local resource admission and selected hosts. Steward is not a second scheduler. AI Hub constructs and advances approved workflows. Steward helps explain the machine's condition while that work runs. Master Suite embeds the three native workspaces but must not create a second authority on a port already owned by another installation.
For cross-network jobs, nodes make outbound HTTPS connections to their coordinator. Do not expose local worker/GPU ports publicly. An offline node, expired consent, unsupported handler, unavailable system RAM and a failed GPU lease are different reasons a step can wait. Inspect the failing layer rather than restarting every service because the dashboard is stale.
7. Remote controls and Windows actions
Remote broker controls require compatible server support, a separate control credential and explicit target-node opt-in. An accepted request is not proof it was applied: read the node's signed acknowledgement. Existing leases and CPU jobs are not killed by a request to pause new admission. Resuming never bypasses local capacity or telemetry checks. Do not blindly resend an unknown operation.
Local administrative actions show the target and require Windows approval for the selected change. Protected system/ASL/recovery/security tasks cannot be disabled through a generic control. Signing out another account can close its apps and lose unsaved work; inspect its purpose before proceeding.
Action history records intent before launching the helper and records the result afterward. Unconfirmed means the helper or Windows operation may still be running. Inspect the exact recorded process identity, not only a recycled PID. Do not force-stop a helper or repeat the operation just because its window is gone.
If action history is damaged, Steward shows a recovery error and blocks another Windows action. It does not skip bad lines, treat the journal as empty or hide an older pending helper outside the most recent page. Preserve the original file. Reconcile live helpers before any repair; never delete the whole AppData folder. The separate ACTION-HISTORY-RECOVERY guide gives the operator-level procedure.
8. Temperatures, fans and cooling controls
Optional AMD adapter: review, accept or decline
Open Temperatures & Cooling → Review AMD component terms before using AMD readings or a new AMD cooling boost. Read the complete scrollable terms; use Save a copy of these terms (or Save terms in Master Suite) for a local copy. Saving or closing does not accept. Choose Accept AMD terms only if you agree and are authorized to do so, or Decline AMD adapter to keep it off. No acceptance is preselected. Purchasing or installing the app is not acceptance.
Only this optional AMD component is affected. NVIDIA readings/controls, Windows diagnostics and fleet work remain usable. Your other hardware need not be AMD. The page shows the recorded terms version and decision time. Steward and Master Suite each retain their own decision; neither silently copies the other's profile. After deciding, refresh explicitly. Acceptance does not start a scan or fan boost.
You can decline later from the same dialog. This blocks new AMD reads and boosts; it does not interrupt restoration already in progress. An explicit request to restore an existing saved AMD cooling profile is not blocked by the consent check. Restoration still requires the exact supported device, a valid protected saved profile and Windows approval; it cannot create a new boost. If controls are unavailable, use the manufacturer's controls and contact support rather than terminate a possibly restoring helper.
The small local consent record contains schema, terms version, text fingerprint, decision and time—no prompts, files, fleet keys or usage telemetry. It is not sent to a server. Missing, damaged or outdated acceptance disables only this adapter until you review the current terms again. The bundled text is checked against its versioned fingerprint before acceptance or use. Full terms: ASL AMD Hardware Adapter Supplemental Terms, version 2026-09-13.1.
Interpreting readings and changing cooling
Read the source label for each value. CPU thermal zones are not necessarily individual core temperatures. NVIDIA fan percentage is a requested duty value, not a measured RPM. Hardware may expose one GPU-wide RPM value rather than a separate sensor per fan. A missing hotspot or fan sensor is not a zero reading.
Availability depends on hardware, firmware, installed drivers and supported vendor interfaces. Steward does not bundle or install a replacement display driver or an unsigned kernel sensor driver. Optional cooling controls appear only when the reviewed adapter reports support.
Cooling boost is a bounded, explicitly approved action, not permanent manual fan tuning. NVIDIA automatic control and AMD's saved fan-curve/Zero RPM settings have different restoration paths. Check the reported restoration result. Do not run competing fan utilities, force-stop the helper or repeatedly boost when restoration is unknown. Follow the hardware vendor's recovery guidance if automatic control cannot be confirmed. See the AMD_COOLING guide for specifics.
Interactive cooling is not a service-account or remote-executor feature. Fleet membership and an AI workflow do not authorize another computer to tune your fans.
9. Safe Cleanup and retained data
Cleanup starts off. First preview eligible aged files in the supported current-user Temp and CrashDumps categories. The minimum age is seven days, with at most 500 files and 512 MiB per run. Changed, locked, linked and unsupported files are skipped. Documents, Downloads, projects, browser profiles, system folders and the Recycle Bin are not automatic cleanup targets.
Confirming cleanup authorizes permanent deletion, not a move to the Recycle Bin. Steward cannot undo it. The separately enabled in-app schedule authorizes future bounded runs for the selected rule while Steward is open; it does not ask again on every scheduled run. Disable the rule or close Steward to stop it. No hidden system-wide cleanup service or AI task is installed.
Local Data controls separately clear the eligible Steward histories and generated reports after confirmation. User-owned exports and files shared elsewhere can remain after uninstalling. Damaged or pending action evidence requires recovery, not blind deletion. Back up what matters before removing retained records.
10. Troubleshooting and support checklist
- GPU broker healthy, Parallel unavailable: these are separate services. Read the worker's version, last contact, ownership and supervisor error record. A quiet error file alone does not prove the service exited cleanly.
- Unmanaged GPU memory: inspect reported process identity, owner, observation time and dedicated allocation. Do not reclaim another application's lease or assume every emulator allocation is unused.
- Fleet authentication failure: verify the intended server and viewer role. Do not borrow another application's credentials or restart active services.
- Partial scan: expand the failing source and reason. Try a deliberate fresh scan after the underlying provider recovers; do not label missing data healthy.
- Unknown action result: preserve its timestamp and exact target, reconcile the helper and Windows state, then decide the next action. Do not repeat blindly.
- Fan restore unknown: stop issuing boosts, preserve the record, and use the supported automatic-control/vendor recovery path.
For help, export a support bundle only if useful, open it and review the contents before sending. Identifying details are minimized by default, not guaranteed anonymous. Steward does not send the bundle automatically. Email support@autosecurelogin.com with the app/Windows/driver versions, exact state, time and timezone, last successful step and a concise description of expected behavior. A product purchase does not imply a guaranteed support response time.