Aller au contenu principal

Build AI Workflows

The Hydra Builder is TietAI's visual workflow designer. You build clinical AI pipelines by placing nodes on a canvas, connecting them, configuring each one, and running the result. This guide covers the full workflow authoring experience from opening the studio to reading execution results.


Accessing the Hydra Builder

In the main left navigation, click Hydra Builder. You land on the Workflows overview page, which lists all pipelines in your organization.

  • To open an existing workflow: click its name
  • To start a new one: click New Pipeline in the top-right corner, give it a name and optional description, then click Create

The canvas opens with the node palette on the left, the toolbar at the top, and the execution panel at the bottom (collapsed by default).


Canvas controls

ActionHow to do it
Pan the canvasClick and drag on empty canvas space
Zoom in / outScroll wheel, or use the + / - buttons in the bottom-right corner
Fit to screenPress Ctrl+Shift+F or click the fit icon in the bottom-right
Select a nodeSingle-click
Select multiple nodesHold Shift and click each node, or drag a selection rectangle
Move a nodeClick and drag the node header
Open node configDouble-click the node
Delete a nodeSelect it, then press Delete or Backspace
Delete a connectionClick the connection line to select it, then press Delete
UndoCtrl+Z
RedoCtrl+Shift+Z or Ctrl+Y
Copy / paste nodesCtrl+C / Ctrl+V (pastes with offset so nodes don't overlap)

The node palette

The palette on the left lists every available node, organized into categories. Use the search bar at the top to find nodes by name.

Category overview

CategoryDescriptionExample nodes
HealthcareFetch and write clinical dataEpic Patient List, Cerner FHIR Query, FHIR Write, FHIR Transform
AI / MLLanguage models and ML inferenceClinical Summarizer, AI Agent, NER Extractor, ML Inference, LLM Processor
DataTransform, filter, and route dataFHIR Transform, Filter, Conditional, Router, JSONPath Transform, Merge
OutputProduce results and send data outReport Generator, Notification, FHIR Create, FHIR Update, API Call
ControlManage execution flowScheduler, Loop, Parallel, Error Handler, Batch
astuce

If you cannot find a node by scrolling, use the search bar. Searching for "fhir" returns all FHIR-related nodes regardless of category.


Adding a node

Method 1 — Drag from palette: Click a node in the palette and drag it to the desired position on the canvas. Release to drop.

Method 2 — Double-click palette item: Double-clicking a node in the palette adds it to the center of the current canvas view.

Once a node is on the canvas, it shows its name, category color, and its input/output port circles on the left and right edges.


Connecting nodes

Ports are the connection points on each node:

  • Input ports appear on the left side (circle outline)
  • Output ports appear on the right side (circle filled)

To draw a connection:

  1. Hover over the source node — the output port highlights
  2. Click and drag from the output port
  3. Drag to the input port on the destination node
  4. Release — the connection line snaps into place

Port compatibility and color coding

Connections are type-checked. You cannot connect a text output to a fhir-bundle input.

Connection colorMeaning
Blue / dark grayValid connection between compatible port types
Red (during drag)Incompatible port types — connection will be rejected

If you attempt an incompatible connection, the canvas prevents it and shows a tooltip explaining the mismatch (e.g., "Output type text is not compatible with input type fhir-bundle").

A single output port can connect to multiple input ports (fan-out), allowing the same data to flow into parallel branches.


Configuring a node

Double-click any node to open its configuration dialog. The dialog is divided into sections:

  • Required fields — Marked with a red asterisk. The workflow cannot run if these are empty.
  • Optional fields — Provide defaults when left blank; filling them gives more control.
  • Advanced settings — Collapsed by default; expand them for timeout, retry, and error handling options.

After filling in the fields, click Save Configuration. The node header updates to reflect key configured values (such as the connector name or operation type).

remarque

Configuration is saved per-node, not per-workflow. Clicking Save on a node saves that node's config immediately — you still need to save the overall workflow with Ctrl+S or the toolbar Save button to persist the pipeline structure.


Running a workflow

Immediate run

Click Run in the top toolbar. A confirmation dialog appears. Click Run now to start immediately.

The execution panel at the bottom of the screen expands automatically and shows live status for each node as the workflow runs.

Scheduled run

Click the dropdown arrow next to the Run button and select Schedule.

Configure the schedule:

  • Frequency — Once, Hourly, Daily, Weekly, Monthly, or Custom (cron expression)
  • Time — The time the workflow should trigger (in your organization's configured timezone)
  • Start date / End date — Optional date range for the schedule

Click Save Schedule. The workflow's status badge changes from Draft to Scheduled.


Reading execution results

Open the Execution Panel (clock icon at the bottom of the screen, or the Executions tab in the workflow detail view).

Each execution row shows:

  • Execution ID — A unique identifier you can reference when contacting support
  • Status — Completed (green), Failed (red), Running (blue spinner), Pending (gray)
  • Triggered by — The user who ran it, or "Scheduler" for automated runs
  • Duration — Total wall-clock time
  • Records processed — How many data items (patients, messages, resources) flowed through

Click any execution row to open the execution detail view:

  • Canvas overlay — Each node shows a colored status badge: green (success), red (error), gray (not reached)
  • Node detail — Click a node badge to see its per-node log, timing, input snapshot, and output snapshot
  • Error tab — If a node failed, the Error tab shows the full error message, stack trace, and suggested remediation

Saving and naming workflows

Workflows auto-save every 30 seconds as you edit. To save manually: Ctrl+S or the Save button in the toolbar.

To rename a workflow: click the workflow name in the toolbar breadcrumb and edit it inline.

Version history: Click the clock icon in the toolbar to see a list of saved versions. Each save creates a version snapshot. Click any version to preview it; click Restore to roll back.


Tips and keyboard shortcuts

Search the node palette — The palette search bar filters by node name and description. Press / to focus the search bar from the canvas.

Align nodes — Select multiple nodes and use the alignment tools in the right-click context menu (Align left, Align top, Distribute horizontally) to keep your canvas tidy.

Group and label — Right-click on empty canvas space to add a Label (a text annotation) or a Group Box (a visual grouping rectangle). These do not affect execution but make complex canvases easier to read.

Duplicate a node — Select a node and press Ctrl+D to create a copy with the same configuration. Useful when you need multiple nodes of the same type with similar settings.

ShortcutAction
Ctrl+SSave workflow
Ctrl+ZUndo
Ctrl+Shift+ZRedo
Ctrl+DDuplicate selected node
DeleteDelete selected node or connection
/Focus palette search
Ctrl+Shift+FFit all nodes to screen
Ctrl+ASelect all nodes