workflow-hooks

README

Workflow Hooks Example

This example demonstrates how to use hooks with workflow agents for monitoring and logging.

Overview

Hooks provide a way to monitor and instrument workflow execution. This example shows how to:

• Add metrics hooks to track execution statistics
• Add logging hooks to capture workflow events
• Use hooks with both sequential and parallel workflows
• Collect and display metrics after workflow completion

Features Demonstrated

1. Metrics Collection: Track requests, errors, and execution times
2. Logging Integration: Capture workflow events and progress
3. Hook Composition: Use multiple hooks simultaneously
4. Workflow Monitoring: Monitor both sequential and parallel workflows

Usage

# Run the example
go run main.go

Examples

1. Sequential Workflow with Hooks

Demonstrates a data processing pipeline with three sequential steps:

• Data Processor (100ms)
• Data Analyzer (150ms)
• Data Formatter (50ms)

Total expected time: ~300ms + overhead

2. Parallel Workflow with Hooks

Demonstrates parallel processing with three concurrent agents:

• Fast Processor (100ms)
• Medium Processor (200ms)
• Slow Processor (300ms)

Total expected time: ~300ms (slowest agent) + overhead

Code Structure

// Create hooks
metricsHook := core.NewLLMMetricsHook()
loggingHook := core.NewLoggingHook()

// Add hooks to workflow
workflow := workflow.NewSequentialAgent("pipeline").
    WithHook(metricsHook).
    WithHook(loggingHook).
    AddAgent(agent1).
    AddAgent(agent2)

// Run workflow and collect metrics
result, err := workflow.Run(ctx, initialState)
metrics := metricsHook.GetMetrics()

Available Hooks

LLMMetricsHook

Tracks execution statistics:

• Total requests
• Total errors
• Total execution time
• Average execution time
• Request/error counts per operation

LoggingHook

Provides structured logging:

• Agent start/completion events
• Error logging with context
• Execution timing information

Metrics Output

After workflow completion, you can access:

metrics := metricsHook.GetMetrics()
fmt.Printf("Total requests: %d\n", metrics.TotalRequests)
fmt.Printf("Total errors: %d\n", metrics.TotalErrors) 
fmt.Printf("Average time: %v\n", metrics.AverageExecutionTime())

Next Steps

• See the conditional workflow example for branching logic
• Explore custom hook implementation for specialized monitoring
• Try combining workflows with tool execution for more complex pipelines