Skip to main content

​
About

Workflows let you design and automate complex, multi-step AI-driven processes. Each workflow is a graph of connected nodes — from LLM-powered conversation steps to tool calls, HTTP requests, EHR integrations, and more — orchestrated by a runtime execution engine that handles state, routing, retries, and dynamic variable substitution.
Workflows are the backbone of Interactly’s agentic capabilities. Whether you’re automating patient intake, appointment scheduling, or multi-turn conversations, workflows give you full control over logic, branching, and integrations.

​
Workflow Types

Agentic Workflow

Conversational, event-driven execution
  • Multi-turn LLM-powered conversations
  • Dynamic branching via conditional edges
  • Real-time tool and API calls
  • Triggered by calls, campaigns, or webhooks

Scheduled Workflow

Time-based automatic execution
  • Schedule a one-time run at a specific future datetime
  • Runs without manual intervention
  • Supports all node and tool types
  • Concurrency-controlled execution

Triggered Workflow

Event-driven external triggers
  • SQS message-based execution
  • Inbound webhook triggers
  • Integrate with external systems
  • Real-time response to platform events

​
Core Concepts

Nodes are the building blocks of a workflow. Each node performs a discrete action — generating an LLM response, calling a tool, sending an SMS, querying an EHR, or controlling conversation flow. Interactly supports 26 node types across categories including LLM, Communication, Healthcare (EHR), HTTP, and Workflow Control.See Node Types for a full reference.
Edges define how execution flows between nodes. There are three edge types:
  • Direct Edge — unconditional transition to the next node
  • Conditional Edge — natural language condition evaluated by an LLM to decide the route
  • Companion Edge — runs a parallel node alongside the primary flow
Tools extend what LLM nodes can do. Attach inbuilt functions, inline Python scripts, external REST APIs, or Knowledge Base queries to any LLM node. Tools are defined in a registry and resolved at runtime with full JSON schema support.See Tools & Integrations for details.
Global Variables are shared key-value pairs accessible across all nodes in a workflow using {{variable_name}} syntax. They support dynamic substitution in prompts, messages, and tool inputs — enabling personalized, context-aware execution at runtime.
Every workflow supports full version history. Changes are tracked as new versions, with a single active version driving live execution. You can view diffs, import/export versions, and roll back at any time without disrupting running workflows.See Workflow Versioning for details.

​
Key Capabilities

Multi-LLM Support

Run nodes on any supported LLM provider — Azure OpenAI, OpenAI, Google, Anthropic, AWS Bedrock, or custom endpoints — with automatic fallback groups if a provider is unavailable.

Versioning & Rollback

Full version history for every workflow. Publish new versions without disrupting active runs. Roll back instantly if needed.

Simulation

Compare two workflow configurations head-to-head across multiple automated runs. Evaluate outputs and scores side by side before activating a new version.

Webhook Lifecycle Events

Subscribe to workflow lifecycle events — created, updated, deleted, run started, run completed — delivered via outbound webhooks with exponential backoff retry.

Tool & MCP Integration

Attach any combination of inbuilt functions, inline Python, external APIs, or Knowledge Bases to LLM nodes. Supports Model Context Protocol (MCP) for dynamic tool discovery.

DND & Callback Aware

Workflows respect Do-Not-Disturb preferences. Post-run analysis can automatically trigger DND entries or schedule callbacks based on conversation outcomes.

​
Workflow Lifecycle

​
Getting Started

1

Build Your Workflow

Navigate to the Workflows tab in the dashboard. Click New Workflow and start adding nodes from the node palette.
2

Connect Nodes with Edges

Link nodes using Direct, Conditional, or Companion edges. For conditional routing, define natural language conditions — the runtime evaluates them via LLM at execution time.
3

Attach Tools & Set Variables

Add tools to LLM nodes for extended capabilities. Define Global Variables for dynamic data like patient names, appointment times, or custom context.
4

Simulate Before Publishing

Use the Simulation feature to compare two workflow configurations across multiple automated runs, evaluate outputs, and confirm your changes perform as expected.
5

Publish & Monitor

Publish your workflow version and monitor execution runs in real-time via the dashboard or webhook events.

​
API Integration

Programmatic Workflow Management

All workflow features available in the dashboard are fully accessible via the Workflows API — create, configure, execute, and monitor workflows programmatically.
Available via API:
  • ✅ Workflow CRUD (create, read, update, delete)
  • ✅ Version management and publishing
  • ✅ Direct workflow execution
  • ✅ Run history and checkpointing
  • ✅ Schedule management
  • ✅ Simulation control
  • ✅ Webhook subscription management

Workflows API

POST /api/workflows Create and configure workflows programmatically with full feature parity to the dashboard

Workflow Execution API

POST /api/workflows/:id/execute Trigger workflow runs directly via API with custom input payloads
Workflow APIs are ideal for embedding agentic processes into your existing applications, CRM systems, or event-driven pipelines.

​
Next Steps

Building a Workflow

Step-by-step guide to creating your first workflow from scratch

Node Types

Full reference for all 26 node types across every category

Tools & Integrations

Attach tools, external APIs, and Knowledge Bases to LLM nodes

Workflow Versioning

Manage versions, publish changes, and roll back safely

Simulation

Test workflows interactively before going live

Workflow Webhooks

Subscribe to lifecycle events for real-time integrations