PF research library · 4 min read
Try PF Simulation in your browser
r34 documentation · hosted-trial.md
On this page
The hosted PF Simulation trial/demo is live. Create your own account to explore the r34 workbench, configure a study, run a calculation and inspect its results. No local installation is needed.
Create an account and sign in
- Open PF Simulation hosted trial. The website's trial link opens a new browser tab or window.
- Choose Create account, enter your details and use a password of at least 12 characters. Follow the verification link sent to your email address.
- Sign in with your own PF Simulation account. Keep account verification, invitation and password-reset links private.
- Select your Organization and Project, then open the workbench. If you joined somebody else's organization, the projects and actions available to you depend on the access they granted.
Run a first study
- Choose a study purpose and starting configuration, then review the physical topology and current inputs before running. The preset selects its natural topology; incompatible conversions are disabled with an explanation. After edits or imports, inspect the effective Fields/JSON inputs rather than relying on a retained preset name.
- Submit the study. A run may wait in the queue until execution capacity is available; keep track of it in Run history.
- Open the completed run to inspect its reports, plots and recorded configuration. Download available outputs or the run ZIP when you want a local copy.
- To create a trajectory comparison, use Load live orbit comparison preset from Configure. This explicitly replaces the current study after confirmation; it does not preserve all mission settings. Export your current configuration first if you need it. Review the new preset before running.
- For a private workflow, open its completed run and explore the ISAES scientific report. See the ISAES guide and private workflow development guide for the broader workflow.
The workbench configuration guide explains effective inputs, live epoch counts and the distinction between scenario simulation and saved orbit-comparison outputs. Every new simulation calculates its own results. Validated reference inputs can be reused within their freshness rules; this does not reuse previous simulation outcomes. An Earth-blocked optical geometry can legitimately produce no received photon pairs.
Use the AI research assistant
Open the Assistant to select a saved conversation. Opening or reopening the page reads your saved workspace; it does not silently import another completed run. Choose Open selected run when you explicitly want to capture that run's evidence in a conversation. Existing attachments remain associated with their saved conversation.
For workflow authoring, reopening the page selects an existing conversation. Choose New authoring conversation to start another. If a read fails, the loading controls are released and the error is shown so you can retry. Credentials remain separate from conversation evidence: a saved provider configuration does not itself prove that the provider is reachable.
The 16 October release consolidates the confirmed Assistant Recovery, Loading and Opening fixes. Shared behavior is consistent across standalone and hosted workbenches, subject to the hosted service's permissions and enabled features. See the Assistant guide, consolidation notes and workflow reliability guide. An existing private-workflow version may require validation again if it was validated against the earlier runtime descriptor; its stored source and history are retained.
Accounts, storage and optional services
Use your own account and projects for trial work. Access to a shared organization does not automatically grant every project or private assistant workspace. The project access guide explains team permissions.
Ordinary simulation use and classical ISAES reporting do not require a personal AI key or personal storage connection. BYOK and BYOS are optional and remain subject to the hosted service's enabled features and policies. External providers may apply their own charges and limits.
The trial uses the service's configured compute, queue and storage allowances. Check Usage & billing and any messages shown when submitting a run. Reaching a limit can delay or prevent new work. Execution capacity is shared; a configured concurrency limit is not a promise of immediate execution, unlimited use or a particular completion time.