Iterative agent execution with convergence conditions.
Loop workflows execute an agent repeatedly until a condition is met or a maximum iteration count is reached.
package main
import (
"context"
"fmt"
"log"
"strings"
"github.com/agenticgokit/agenticgokit/v1beta"
)
func main() {
// Create refining agent
refiner, err := v1beta.NewBuilder("Refiner").
WithConfig(&v1beta.Config{
LLM: v1beta.LLMConfig{Provider: "openai", Model: "gpt-4"},
}).
Build()
if err != nil {
log.Fatal(err)
}
// Convergence function - stops when quality threshold met
converged := func(iteration int, result *v1beta.Result) bool {
// Stop after 5 iterations max
if iteration >= 5 {
return true
}
// Stop if result contains "FINAL"
return strings.Contains(result.Content, "FINAL")
}
// Create loop workflow
workflow, err := v1beta.NewLoopWorkflow(
"RefinementLoop",
refiner,
converged,
v1beta.WithMaxIterations(5),
v1beta.WithLoopPrompt("Refine the following text. Add 'FINAL' when perfect"),
)
if err != nil {
log.Fatal(err)
}
// Execute loop
results, err := workflow.Run(context.Background(), "Write a short Go tutorial")
if err != nil {
log.Fatal(err)
}
// Display iterations
fmt.Printf("Converged after %d iterations\n\n", len(results))
for i, result := range results {
fmt.Printf("=== Iteration %d ===\n", i+1)
fmt.Println(result.Content)
fmt.Println()
}
}type ConvergenceFunc func(iteration int, result *v1beta.Result) bool
// Simple iteration limit
converged := func(iteration int, result *v1beta.Result) bool {
return iteration >= 10
}
// Content-based convergence
converged := func(iteration int, result *v1beta.Result) bool {
return strings.Contains(result.Content, "COMPLETE")
}
// Metadata-based convergence
converged := func(iteration int, result *v1beta.Result) bool {
if score, ok := result.Metadata["quality_score"].(float64); ok {
return score > 0.95
}
return false
}workflow, _ := v1beta.NewLoopWorkflow(
"Loop",
agent,
converged,
v1beta.WithMaxIterations(10), // Max iterations
v1beta.WithLoopPrompt("Improve the text"), // Prompt for each iteration
v1beta.WithLoopTimeout(5*time.Minute), // Total timeout
)reviewer, _ := v1beta.NewBuilder("CodeReviewer").WithConfig(&v1beta.Config{
LLM: v1beta.LLMConfig{Provider: "openai", Model: "gpt-4"},
}).Build()
converged := func(iteration int, result *v1beta.Result) bool {
// Stop if no more issues found
return !strings.Contains(result.Content, "ISSUE:")
}
workflow, _ := v1beta.NewLoopWorkflow(
"CodeReview",
reviewer,
converged,
v1beta.WithMaxIterations(3),
v1beta.WithLoopPrompt("Review code and suggest improvements"),
)researcher, _ := v1beta.NewBuilder("Researcher").WithConfig(&v1beta.Config{
LLM: v1beta.LLMConfig{Provider: "openai", Model: "gpt-4"},
}).Build()
converged := func(iteration int, result *v1beta.Result) bool {
// Stop when comprehensive
wordCount := len(strings.Fields(result.Content))
return wordCount > 500 && strings.Contains(result.Content, "Conclusion")
}
workflow, _ := v1beta.NewLoopWorkflow(
"Research",
researcher,
converged,
v1beta.WithMaxIterations(5),
)editor, _ := v1beta.NewBuilder("Editor").WithConfig(&v1beta.Config{
LLM: v1beta.LLMConfig{Provider: "openai", Model: "gpt-4"},
}).Build()
converged := func(iteration int, result *v1beta.Result) bool {
// Stop when polished
hasIntro := strings.Contains(result.Content, "Introduction")
hasConclusion := strings.Contains(result.Content, "Conclusion")
return iteration >= 3 && hasIntro && hasConclusion
}
workflow, _ := v1beta.NewLoopWorkflow(
"ContentRefinement",
editor,
converged,
v1beta.WithLoopPrompt("Improve structure and clarity"),
)type LoopState struct {
improvements []string
scores []float64
}
state := &LoopState{}
converged := func(iteration int, result *v1beta.Result) bool {
// Track improvements
state.improvements = append(state.improvements, result.Content)
// Calculate quality score (simplified)
score := float64(len(result.Content)) / 1000.0
state.scores = append(state.scores, score)
// Converge if score plateaus
if len(state.scores) >= 3 {
recent := state.scores[len(state.scores)-3:]
variance := calculateVariance(recent)
return variance < 0.01 // Low variance = plateau
}
return false
}var previousFeedback []string
converged := func(iteration int, result *v1beta.Result) bool {
// Accumulate feedback
previousFeedback = append(previousFeedback, result.Content)
// Inject previous iterations into next prompt
if iteration < 5 {
// Modify agent prompt to include history
return false
}
return true
}results, err := workflow.Run(ctx, input)
if err != nil {
if loopErr, ok := err.(*v1beta.LoopWorkflowError); ok {
fmt.Printf("Failed at iteration %d: %v\n", loopErr.Iteration, loopErr.Err)
// Access results from successful iterations
fmt.Printf("Completed %d iterations successfully\n", len(loopErr.PartialResults))
for i, result := range loopErr.PartialResults {
fmt.Printf("Iteration %d: %s\n", i+1, result.Content)
}
}
}// Prevent infinite loops
workflow, _ := v1beta.NewLoopWorkflow(
"Loop",
agent,
converged,
v1beta.WithMaxIterations(10), // Hard limit
v1beta.WithLoopTimeout(5*time.Minute), // Time limit
v1beta.WithIterationTimeout(30*time.Second), // Per-iteration limit
)converged := func(iteration int, result *v1beta.Result) bool {
// Immediate termination conditions
if strings.Contains(result.Content, "ERROR") {
return true
}
if strings.Contains(result.Content, "CANNOT_IMPROVE") {
return true
}
// Normal convergence check
return iteration >= 10
}go get github.com/agenticgokit/agenticgokit/v1beta
export OPENAI_API_KEY="sk-..."
go run main.go- Subworkflows - Nested loop workflows
- DAG Workflow - Complex dependencies
- Custom Handlers - Custom loop logic
- Workflows Guide - Complete workflow documentation
- Performance - Loop optimization strategies
- Error Handling - Handling loop errors