Hydra Builder
Hydra Builder is TietAI's visual pipeline designer. It provides a drag-and-drop canvas for building, testing, and monitoring clinical AI pipelines — without writing code. The Builder is powered by the Hydra Platform engine: a Go-native, HIPAA/GDPR-compliant agentic runtime with 128+ registered node handlers and support for FHIR R4, HL7v2, DICOM, X12 EDI, and OMOP CDM.
What Hydra Builder is
Hydra Builder is where you build pipelines: ordered graphs of processing nodes connected by typed data ports. A pipeline might fetch patient data from Epic, send it through an AI summarizer, validate the result as a FHIR bundle, and write it to your FHIR server — all in a visual, traceable, repeatable workflow.
Pipelines built in Hydra Builder are executed by the Hydra Platform backend engine, which handles scheduling, retries, error logging, distributed tracing, and execution history.
Hydra Builder also includes the Hydra Copilot — an AI assistant embedded in the canvas that can design pipelines from natural language, diagnose execution errors, and apply multi-step fixes automatically.
Layout overview
The Builder window has five regions:
| Region | Location | Purpose |
|---|---|---|
| Toolbar | Top | New, Save, Run, Schedule, Version History, Settings, Share |
| Node palette | Left panel | Browse and search all available nodes |
| Canvas | Center | Drag, connect, and arrange nodes |
| Properties panel | Right panel | Configure the selected node (appears when a node is selected) |
| Execution panel | Bottom | View past and current runs; inspect per-node results |
Toolbar buttons
| Button | Shortcut | Action |
|---|---|---|
| New | — | Creates a new pipeline in a new tab |
| Save | Ctrl+S | Saves the current canvas state as a new version |
| Run | — | Executes the pipeline immediately |
| Run ▾ (dropdown) | — | Opens the schedule dialog to set a recurring run |
| Pause | — | Suspends a running execution (appears only when a pipeline is running) |
| Version History | — | Opens the version timeline; click any version to restore |
| Settings | — | Pipeline-level settings: timeout, max retries, tags, description |
| Share | — | Share this pipeline with specific users or teams (read or edit access) |
Canvas interactions
| Action | Method | Keyboard shortcut |
|---|---|---|
| Pan canvas | Click and drag empty space | Arrow keys (slow pan) |
| Zoom in | Scroll up | + |
| Zoom out | Scroll down | - |
| Fit to screen | Click fit icon (bottom-right) | Ctrl+Shift+F |
| Select node | Single-click | — |
| Multi-select | Shift+click or drag selection box | Ctrl+A (select all) |
| Move node | Drag node header | — |
| Open config | Double-click node | Enter when node is selected |
| Delete node | Select → press Delete | Delete or Backspace |
| Delete connection | Click connection line → Delete | — |
| Undo | — | Ctrl+Z |
| Redo | — | Ctrl+Shift+Z |
| Duplicate node | — | Ctrl+D |
| Copy / Paste | — | Ctrl+C / Ctrl+V |
| Focus palette search | — | / |
Node palette
The left panel shows all available nodes. Use the search bar at the top to find nodes by name or keyword. Nodes are grouped into categories:
Healthcare
Nodes that interact with clinical systems and health data standards.
| Node | What it does |
|---|---|
| Epic FHIR | Fetches patient data from Epic via FHIR R4 and Epic-specific APIs |
| Cerner FHIR | Fetches FHIR resources from Cerner / Oracle Health |
| Allscripts FHIR | Connects to Allscripts Unity and TouchWorks APIs |
| Dedalus FHIR | GDPR/ENS-compliant connector for Dedalus systems |
| FHIR Query / Write / Transform | Generic FHIR R4 read, write, and transform operations |
| HL7 Listener | Receive inbound HL7 v2 messages via MLLP |
| HL7 to FHIR | Convert HL7 v2 messages to FHIR R4 resources |
| DICOM Query / Retrieve | Query and retrieve medical imaging from DICOMweb servers |
| SAP Patient Management | ADT workflows for SAP IS-H |
| SAP Clinical Data | Orders, results, and clinical documentation from SAP IS-H*MED |
| athenahealth FHIR | Fetches FHIR resources from athenahealth |
| EMPI Lookup | Queries the Enterprise Master Patient Index for patient matching and linking |
AI / ML
Nodes that apply AI or machine learning to clinical data.
| Node | What it does |
|---|---|
| AI Agent | Multi-step reasoning agent with optional tool use |
| Hydra Copilot | AI assistant that designs and fixes pipelines from natural language |
| Clinical Summarizer | Summarizes patient clinical history from FHIR data |
| NER Extractor | Extracts named entities (diagnoses, medications, labs) from clinical text |
| LLM Processor | Applies a custom prompt to input text using a language model |
| ML Inference | Runs a custom ML model endpoint on structured input data |
| Clinical Decision Support | Rule-based and AI-based clinical decision support |
| Risk Score | Computes patient risk scores (readmission, deterioration, etc.) |
| EMPI Lookup | Queries the Enterprise Master Patient Index to match, merge, or link patient records across systems |
| Europe PMC Search | Searches biomedical literature from Europe PMC for evidence-based clinical context |
Data
Nodes for transforming, filtering, and routing data within a pipeline.
| Node | What it does |
|---|---|
| Filter | Keeps only records that match a condition |
| Conditional | Branches the pipeline based on a data condition |
| Router | Routes data to different downstream paths based on rules |
| JSONPath Transform | Extracts or reshapes JSON fields using JSONPath expressions |
| Merge | Combines multiple upstream inputs into one |
| Batch | Groups records into batches for downstream processing |
| Aggregate | Reduces a list of records to a summary value |
| Sort | Sorts records by a specified field |
Output
Nodes that produce results or write data to external systems.
| Node | What it does |
|---|---|
| Report Generator | Produces a PDF or CSV report from pipeline output |
| Notification | Sends an in-app notification or email to specified users |
| API Call | Makes an outbound HTTP request to an external API |
| Database Write | Writes structured records to a connected database |
| X12 Generate | Generates X12 EDI transactions (claims, eligibility) |
Claims & Billing
Nodes for healthcare claims processing and revenue cycle management.
| Node | What it does |
|---|---|
| X12 Parse | Parses inbound X12 EDI transactions (835, 837, 270/271, 276/277, 278) |
| X12 Generate | Generates outbound X12 EDI transactions for claims submission |
| X12 Validate | Validates X12 transactions against HIPAA implementation guides |
| X12 to FHIR | Converts X12 transactions to FHIR ExplanationOfBenefit and Claim resources |
| FHIR to X12 | Converts FHIR Claim resources to X12 837 transactions |
Control
Nodes that manage execution flow.
| Node | What it does |
|---|---|
| Scheduler | Triggers the pipeline on a schedule (used as an input node) |
| Loop | Iterates over a list and runs downstream nodes for each item |
| Parallel | Runs multiple branches simultaneously |
| Error Handler | Catches errors from upstream nodes and executes a fallback path |
Hydra Copilot
The Hydra Copilot is an AI assistant embedded in the canvas. Open it by clicking the Copilot button in the toolbar or pressing Ctrl+K.
What the Copilot can do:
- Design a pipeline from text — Describe what you want ("receive HL7 ADT messages, validate, transform to FHIR, write to server") and the Copilot builds the pipeline with correctly wired nodes.
- Diagnose execution errors — When a pipeline run fails, paste or click the error and the Copilot identifies the root cause.
- Auto-fix — The Copilot can apply multi-step fixes directly to the canvas: inserting missing transform nodes, correcting port connections, and adjusting configuration — all without user intervention.
- Explain nodes — Ask the Copilot what any node does, what its output looks like, or which node to use for a given task.
The Copilot reads the full current state of your canvas in each message, so it always has accurate context about the pipeline structure.
Real-time thinking and reasoning
When the Copilot or any AI Agent node processes a request, TietAI streams the response in real time — including intermediate thinking and reasoning steps. In the agent chat interface and the pipeline test panel, you can see:
- Thinking indicators — A live indicator shows when the model is reasoning through a complex problem
- Reasoning trace — Intermediate reasoning steps appear as they are generated, giving visibility into how the agent arrived at its answer
- Streaming output — Final text streams token-by-token rather than appearing all at once
This transparency helps clinicians and pipeline builders understand and trust the AI's decision-making process. Toggle the thinking/reasoning visibility on or off using the eye icon in the agent chat toolbar.
The Copilot works best when you describe the data flow you want — source system, what should happen to the data, and where it should go — rather than asking about specific node names.
Connection rules
Connections are type-checked. The canvas prevents invalid connections and shows a tooltip explaining the mismatch.
| Output type | Compatible input types |
|---|---|
fhir-bundle | fhir-bundle |
fhir-resource | fhir-resource, fhir-bundle (auto-wrapped) |
text | text, json (auto-serialized) |
json | json |
hl7-message | hl7-message |
trigger-event | trigger-event |
any | Any type (used by utility nodes like Router and Error Handler) |
A single output port can fan out to multiple input ports. Each downstream node receives an independent copy of the data.
Workflow states
| State | What it means |
|---|---|
| Draft | Pipeline has been created and edited but not yet run |
| Running | At least one execution is currently in progress |
| Scheduled | Has a schedule configured; will run automatically |
| Completed | Last execution finished successfully (pipeline is idle) |
| Failed | Last execution encountered an error |
| Paused | An execution is paused (manual pause or hit pause point) |
Version history
Every time you save, TietAI creates a version snapshot. To view version history:
- Click the Version History button (clock icon) in the toolbar
- A side panel opens showing saved versions, each with a timestamp and author
- Click any version to preview the canvas at that point in time
- Click Restore this version to roll back — this creates a new version, preserving history
Add meaningful version labels by clicking the pencil icon next to a version entry — for example: "Added FHIR validation" or "Switched Epic endpoint to production".
Sharing a pipeline
Click Share in the toolbar.
- Share with a user — Enter a team member's name. Choose Can view or Can edit.
- Share with a team — Apply access to an entire user group.
- Transfer ownership — (Admins only) Reassign the pipeline to another user.
Pipeline limits
| Limit | Default value |
|---|---|
| Maximum nodes per pipeline | 100 |
| Maximum connections per pipeline | 200 |
| Execution timeout | 30 minutes |
| Maximum concurrent executions per pipeline | 5 |
| Maximum records per batch node | 10,000 |
| Execution history retained | 90 days |
Contact TietAI support to request limit increases for your organization.