Docs /Getting Started
AI Pipelines — Reusable Workflows
When you ask the assistant to perform a complex, multi-step operation — for example a full CFD study from a CT scan — it does not improvise the steps from scratch every time. Instead it can recognize a pipeline: a saved, versioned, and reproducible sequence of operations. Running a pipeline replays a frozen recipe, so the same request produces the same steps every time, with no drift between runs.
This makes complex workflows repeatable (identical execution on every run), reusable (change only the parameters that differ), and auditable (every run is recorded).
Where pipelines come from: LeaRes ships with a set of ready-to-use global pipelines covering the standard medical and engineering workflows. Any multi-step operation you run can also be saved as your own pipeline for later reuse.
Opening the Pipelines panel
Open the chatbot from the bottom-right corner, then click the Pipelines button in the assistant’s shortcut bar (next to Manage memory and Show me the tutorial).
The panel lists every pipeline available to you:
- pipelines tagged
[global]— provided with the platform, available to all users; - pipelines in your own namespace — workflows you saved yourself.
Each entry shows its name and a short description of what it does.
Inspecting a pipeline
Click any pipeline to expand its detail. You will see:
| Section | What it tells you |
|---|---|
| Version & status | e.g. v1.0.0 (published) — the exact, immutable recipe being used |
| Customizable parameters | the typed “knobs” you can change (name, type, and default value) |
| Steps | the ordered list of operations the pipeline runs |
| Recent runs | the history of executions, each with its outcome |
For example, the global Simulate flow (end-to-end CFD) pipeline exposes parameters such as patient_name, organ (default aorta), scan_id, material (default blood), inlet_velocity, mesh_resolution, and duration_seconds, and runs nine steps from load_ct_scan through to show_simulation_results.
The ready-made global pipelines
| Pipeline | What it does |
|---|---|
| Prepare anatomy | Load CT → segment an organ → create a CFD project |
| Configure simulation | Assign material → auto-zone → set boundary conditions |
| Mesh and simulate | Discretize the mesh → run the CFD simulation |
| Simulate flow (end-to-end CFD) | The complete chain: from CT scan to simulated blood/air flow results |
These mirror the composable workflow stages described in the previous chapter — now as named, version-pinned recipes you can reuse.
Using pipelines through the chatbot
You don’t have to open the panel: the assistant works with pipelines directly from natural language.
1. Reuse — let the assistant find a matching pipeline
Before planning a multi-step task, the assistant looks for an existing pipeline that already does it.
| What you can say | What happens |
|---|---|
| ”Is there a pipeline to run an end-to-end CFD from a CT scan?” | The assistant lists matching pipelines with their parameters |
| ”Find a pipeline to segment and reconstruct an organ” | Suggests the closest existing workflow to reuse |
When a match is found, the assistant proposes it and changes only the parameters that differ from the defaults — no re-planning.
2. Run — instantiate with minimal overrides
| What you can say | What happens |
|---|---|
| ”Run the simulate-flow pipeline for patient Mario, aorta” | Executes the frozen recipe, overriding only patient_name and organ |
| ”Run it again but with fine resolution” | Re-runs the same pipeline, changing only mesh_resolution |
Because execution replays a pinned spec, repeated runs do not diverge. Each run is saved under Recent runs with its outcome.
3. Save — capture a new workflow for reuse
After you complete a new multi-step operation, you can turn it into a reusable pipeline.
| What you can say | What happens |
|---|---|
| ”Save these steps as a pipeline called Carotid study” | Records the operation as a reusable draft in your namespace |
Saved pipelines appear in the Pipelines panel and can be run later with the same one-line commands.
Why this matters
- Reproducibility — a pipeline is a frozen, versioned recipe; the same request always runs the same steps.
- Minimal customization — you change only the parameters that differ, not the whole workflow.
- Traceability — every execution is recorded as a run with its result, so you can review what was done.
Module availability: pipelines respect the same module gating as individual commands. A pipeline whose steps require Segmentation, Respirology, or Engineering is only usable when the corresponding module is enabled for your account.