1. Executing & Monitoring Tasks
Once a flow is created, it is crucial to ensure that tasks execute correctly and reach their expected outcomes. Execution monitoring helps track task performance, detect errors, and fine-tune processes for efficiency.
Lifecycle of Task Execution
A task within a node follows a structured execution cycle:
Trigger Activation – The node receives a trigger to begin execution.
Data Processing – Inputs are retrieved and processed according to predefined configurations.
Execution Logic – The agent follows decision-making logic, such as conditional branches or sequential execution.
Action Execution – The agent performs the task (e.g., sending an email, updating a database, generating a response).
Output Handling – The output is processed and passed to the next step or external system.
Completion & Logging – The system logs task completion status, including success, failure, or required intervention.
Monitoring Task Execution
A task within a node follows a structured execution cycle:
Task History – View past executions to identify patterns and troubleshoot recurring issues.
Real-Time Status Indicators – Observe running tasks and pending actions.
How Nodes Process Inputs & Outputs
Each node in a flow requires structured inputs to execute correctly and produce expected outputs.
Input Handling in Nodes
The Agent Dashboard acts as the central hub for managing all agents and performing real-time actions
User Input – Information provided manually by a user.
Stored Data – Data retrieved from a memory source, database, or prior task.
Integration Data – Information pulled from an external system (e.g., Gmail, Slack, CRM).
Generated Data – AI-generated content based on the agent’s prompt and logic.
Output Processing
After executing, nodes produce outputs that serve as:
Direct Responses – Final outputs sent to users or external systems.
Intermediate Data – Passed to the next node for further processing.
Conditional Triggers – Used to determine the next path in a branching flow.
Setting Up & Managing Variables
Beam AI allows defining variables to streamline workflow execution. Variables are used to store and manipulate data dynamically during task execution.
Types of Variables in Beam AI
AI Fill – The system determines the value based on available context.
User Fill – The user manually inputs a value when required.
Static Values – Predefined values that do not change.
Types of Variables in Beam AI
Handling Task Dependencies & Execution Order
Many workflows require tasks to execute in a specific sequence to ensure logical processing. Managing dependencies ensures data is available at the correct time.
Types of Dependencies
✔ Sequential Execution – Each step must complete before the next starts.
✔ Conditional Execution – Execution depends on an event, such as data availability or decision logic.
Example: Task Dependency in an Order Processing Flow
Utilising Database Sources as Inputs
Nodes can pull data from external databases, stored memory, or API calls to enrich execution accuracy.
Example: Integrating a Customer Database
Retrieve Customer Details → Uses stored memory to pull customer order history.
Validate Purchase → Checks order ID against the company database.
Determine Eligibility → Compares against stored return policies.
Error Handling & Debugging in Execution
Error detection and handling ensure workflows remain functional even when unexpected issues arise.
Common Errors & Solutions
Debugging Strategies
Use Execution Logs – Analyse logs to pinpoint failures.
Run Test Executions – Simulate different inputs and check outputs.
Modify Node Settings – Adjust variable configurations or prompts.
Node Optimisation & Execution Accuracy
What is Node Optimisation?
Beam AI introduces Node Optimisation to improve:
✅ Task execution accuracy
✅ Workflow efficiency
✅ Output validation
How it Works
1️⃣ Reviewing Past Executions
Users can navigate to the Tools Page and view past workflow executions.
Each execution is assigned an accuracy score based on output correctness.
2️⃣ Providing Feedback & Refining Outputs
Users can flag errors such as:
Data loss in execution
Missing task inputs
Incorrect memory lookup
Hallucinations (inaccurate AI-generated data)
3️⃣ Optimising & Re-Executing Workflows
Users can apply recommended optimisations and re-run workflows.
Ensures continuous improvement in automated data extraction.