Skip to main content
Workflows enable you to build complex automation through an intuitive Canvas-style visual interface. Drag, connect, and configure nodes to create powerful AI-driven processes.

Overview

AeonSage Workflows provide a visual, node-based approach to automation:
  • Drag-and-drop canvas interface - Build flows visually
  • Connect nodes - Define flow logic
  • Configure nodes - Visual forms for each node
  • Real-time execution - See your workflow run
  • Share workflows - Version as code

Canvas Interface

Visual Flow Builder

┌─────────────────────────────────────────────────────────────────────────────┐
│  Workflow Canvas                                                             │
│                                                                              │
│    ┌─────────┐        ┌─────────┐        ┌─────────┐                       │
│    │ Trigger │───────▶│  Action │───────▶│  Output │                       │
│    │         │        │         │        │         │                       │
│    │ Message │        │ AI Call │        │  Reply  │                       │
│    │ Received│        │ GPT-4o  │        │ Channel │                       │
│    └─────────┘        └─────────┘        └─────────┘                       │
│         │                   │                                                │
│         │              ┌────▼────┐                                          │
│         │              │Condition│                                          │
│         │              │  Check  │                                          │
│         │              └────┬────┘                                          │
│         │                   │                                                │
│         ▼              ┌────▼────┐     ┌─────────┐                          │
│    ┌─────────┐         │  True   │────▶│  Skill  │                          │
│    │  Log    │         │  Path   │     │  Web    │                          │
│    │ Event   │         └─────────┘     └─────────┘                          │
│    └─────────┘                                                               │
│                                                                              │
└─────────────────────────────────────────────────────────────────────────────┘

Canvas Components

Triggers - Events that start a workflow
  • Message Received
  • Schedule
  • Webhook
  • Manual
Actions - Operations performed in the flow
  • AI Call
  • Send Message
  • Run Skill
  • HTTP Request
Logic - Control flow and branching
  • Condition
  • Switch
  • Loop
  • Parallel
Outputs - Results and responses
  • Reply
  • Store
  • Notify
  • Transform

Node Types

Trigger Nodes

Message Received

Fires when a message arrives on configured channels. Configuration:
  • Channels: Select Telegram, Discord, WhatsApp, etc.
  • Filter: Optional expression to filter messages

Schedule

Fires on a time-based schedule. Configuration:
  • Cron expression: e.g., 0 9 * * *
  • Timezone: UTC or Local

Webhook

Fires when external service calls the webhook URL. Configuration:
  • Method: GET or POST
  • Authentication: Enable/disable

Action Nodes

AI Call

Call an AI model with a prompt. Configuration:
  • Provider: OpenAI, Anthropic, Ollama
  • Model: Select available model
  • Prompt: Template with variables
  • Temperature: 0-2 (default 0.7)

Run Skill

Execute a skill action. Configuration:
  • Skill: Select from available skills
  • Action: Select skill action
  • Parameters: JSON configuration

HTTP Request

Make an HTTP request to external API. Configuration:
  • Method: GET, POST, PUT, DELETE
  • URL: Target endpoint
  • Headers: JSON object
  • Body: JSON payload

Logic Nodes

Condition

Branch based on a condition. Configuration:
  • Expression: Boolean expression
  • Outputs: True/False paths

Switch

Multiple branch selection. Configuration:
  • Expression: Value to switch on
  • Cases: Array of case values

Parallel

Execute multiple branches simultaneously. Configuration:
  • Branches: 2-10 parallel branches

Loop

Iterate over array items. Configuration:
  • Array: Expression returning array
  • Max Iterations: Safety limit

Example Workflows

1. Simple Q&A Bot

Flow:
  1. Message Received (Trigger)
  2. Check if message contains ”?” (Condition)
  3. If yes → Call AI with GPT-4o (Action)
  4. Send Reply (Output)
Stats:
  • Nodes: 4
  • Avg Latency: 1.2s
  • Success Rate: 99.5%

2. Multi-Channel Notification

Flow:
  1. Webhook (Trigger)
  2. Transform Data (Action)
  3. Parallel Send to 3 channels (Logic)
  4. Send to Telegram, Discord, Slack (Outputs)
Stats:
  • Nodes: 5
  • Avg Latency: 0.8s
  • Success Rate: 99.9%

3. AI Research Assistant

Flow:
  1. Message: !research (Trigger)
  2. Web Search (Action)
  3. Extract Content (Action)
  4. Loop through sources (Logic)
  5. Summarize with Claude (Action)
  6. Compile Report (Action)
  7. Send Report (Output)
Stats:
  • Nodes: 7
  • Avg Latency: 15s
  • Success Rate: 97%

Workflow as Code

Export and Version

# workflow.yaml
name: "Customer Support Bot"
version: "1.2.0"
trigger:
  type: message
  filter: "channel === 'telegram'"

nodes:
  - id: classify
    type: ai_call
    provider: openai
    model: gpt-4o-mini
    prompt: |
      Classify this message: {{message.text}}
      Categories: question, complaint, feedback

  - id: route
    type: switch
    expression: "{{classify.category}}"
    cases:
      question: handle_question
      complaint: handle_complaint
      feedback: handle_feedback

  - id: handle_question
    type: ai_call
    provider: openai
    model: gpt-4o
    prompt: "Answer: {{message.text}}"

  - id: handle_complaint
    type: parallel
    branches:
      - notify_team
      - acknowledge

connections:
  - from: trigger
    to: classify
  - from: classify
    to: route
  - from: route.handle_question
    to: handle_question

Best Practices

Keep Workflows Focused

Each workflow should have a single, clear purpose. Split complex workflows into smaller, reusable ones.

Handle Errors Gracefully

Add error handling branches to catch failures and provide fallback responses.

Use Variables Wisely

Store intermediate results in variables to avoid redundant AI calls.

Test Thoroughly

Use the debug mode to test workflows step-by-step before deployment.

Next Steps

Workflows provide a powerful visual approach to automation. To get started:
  • Review the node types and configuration options above
  • Study the example workflows for common patterns
  • Experiment with the workflow as code format for version control