feat: make rate limit delay configurable via RATE_LIMIT_DELAY_MS env var

- Add getRateLimitDelay() function to read delay from environment
- When RATE_LIMIT_DELAY_MS is not set or 0, use original model without wrapper
- Remove debug print statements from rateLimitedModel

Change-Id: I491d061b90530a84976e26952c7245f821553d1c

adk/human-in-the-loop/5_supervisor/agent.go

@@ -19,6 +19,8 @@ package main
 import (
 	"context"
 	"fmt"
+	"os"
+	"strconv"
 	"time"
 
 	"github.com/cloudwego/eino/adk"
@@ -47,21 +49,35 @@ func (r *rateLimitedModel) WithTools(tools []*schema.ToolInfo) (model.ToolCallin
 }
 
 func (r *rateLimitedModel) Generate(ctx context.Context, input []*schema.Message, opts ...model.Option) (*schema.Message, error) {
-	fmt.Println("enter rateLimitedModel.generate")
 	time.Sleep(r.delay)
 	return r.m.Generate(ctx, input, opts...)
 }
 
 func (r *rateLimitedModel) Stream(ctx context.Context, input []*schema.Message, opts ...model.Option) (*schema.StreamReader[*schema.Message], error) {
-	fmt.Println("enter rateLimitedModel.stream")
 	time.Sleep(r.delay)
 	return r.m.Stream(ctx, input, opts...)
 }
 
+func getRateLimitDelay() time.Duration {
+	delayMs := os.Getenv("RATE_LIMIT_DELAY_MS")
+	if delayMs == "" {
+		return 0
+	}
+	ms, err := strconv.Atoi(delayMs)
+	if err != nil {
+		return 0
+	}
+	return time.Duration(ms) * time.Millisecond
+}
+
 func newRateLimitedModel() model.ToolCallingChatModel {
+	delay := getRateLimitDelay()
+	if delay == 0 {
+		return commonModel.NewChatModel()
+	}
 	return &rateLimitedModel{
 		m:     commonModel.NewChatModel(),
-		delay: 300 * time.Millisecond,
+		delay: delay,
 	}
 }
 

adk/human-in-the-loop/6_plan-execute-replan/agent.go

@@ -19,6 +19,8 @@ package main
 import (
 	"context"
 	"fmt"
+	"os"
+	"strconv"
 	"strings"
 	"time"
 
@@ -46,21 +48,35 @@ func (r *rateLimitedModel) WithTools(tools []*schema.ToolInfo) (model.ToolCallin
 }
 
 func (r *rateLimitedModel) Generate(ctx context.Context, input []*schema.Message, opts ...model.Option) (*schema.Message, error) {
-	fmt.Println("enter rateLimitedModel.generate")
 	time.Sleep(r.delay)
 	return r.m.Generate(ctx, input, opts...)
 }
 
 func (r *rateLimitedModel) Stream(ctx context.Context, input []*schema.Message, opts ...model.Option) (*schema.StreamReader[*schema.Message], error) {
-	fmt.Println("enter rateLimitedModel.stream")
 	time.Sleep(r.delay)
 	return r.m.Stream(ctx, input, opts...)
 }
 
+func getRateLimitDelay() time.Duration {
+	delayMs := os.Getenv("RATE_LIMIT_DELAY_MS")
+	if delayMs == "" {
+		return 0
+	}
+	ms, err := strconv.Atoi(delayMs)
+	if err != nil {
+		return 0
+	}
+	return time.Duration(ms) * time.Millisecond
+}
+
 func newRateLimitedModel() model.ToolCallingChatModel {
+	delay := getRateLimitDelay()
+	if delay == 0 {
+		return commonModel.NewChatModel()
+	}
 	return &rateLimitedModel{
 		m:     commonModel.NewChatModel(),
-		delay: 300 * time.Millisecond,
+		delay: delay,
 	}
 }
 

adk/human-in-the-loop/7_deep-agents/agent.go

@@ -18,7 +18,8 @@ package main
 
 import (
 	"context"
-	"fmt"
+	"os"
+	"strconv"
 	"time"
 
 	"github.com/cloudwego/eino/adk"
@@ -47,21 +48,35 @@ func (r *rateLimitedModel) WithTools(tools []*schema.ToolInfo) (model.ToolCallin
 }
 
 func (r *rateLimitedModel) Generate(ctx context.Context, input []*schema.Message, opts ...model.Option) (*schema.Message, error) {
-	fmt.Println("enter rateLimitedModel.generate")
 	time.Sleep(r.delay)
 	return r.m.Generate(ctx, input, opts...)
 }
 
 func (r *rateLimitedModel) Stream(ctx context.Context, input []*schema.Message, opts ...model.Option) (*schema.StreamReader[*schema.Message], error) {
-	fmt.Println("enter rateLimitedModel.stream")
 	time.Sleep(r.delay)
 	return r.m.Stream(ctx, input, opts...)
 }
 
+func getRateLimitDelay() time.Duration {
+	delayMs := os.Getenv("RATE_LIMIT_DELAY_MS")
+	if delayMs == "" {
+		return 0
+	}
+	ms, err := strconv.Atoi(delayMs)
+	if err != nil {
+		return 0
+	}
+	return time.Duration(ms) * time.Millisecond
+}
+
 func newRateLimitedModel() model.ToolCallingChatModel {
+	delay := getRateLimitDelay()
+	if delay == 0 {
+		return commonModel.NewChatModel()
+	}
 	return &rateLimitedModel{
 		m:     commonModel.NewChatModel(),
-		delay: time.Minute,
+		delay: delay,
 	}
 }
 

adk/human-in-the-loop/8_supervisor-plan-execute/agent.go

@@ -19,6 +19,8 @@ package main
 import (
 	"context"
 	"fmt"
+	"os"
+	"strconv"
 	"strings"
 	"time"
 
@@ -56,10 +58,26 @@ func (r *rateLimitedModel) Stream(ctx context.Context, input []*schema.Message,
 	return r.m.Stream(ctx, input, opts...)
 }
 
+func getRateLimitDelay() time.Duration {
+	delayMs := os.Getenv("RATE_LIMIT_DELAY_MS")
+	if delayMs == "" {
+		return 0
+	}
+	ms, err := strconv.Atoi(delayMs)
+	if err != nil {
+		return 0
+	}
+	return time.Duration(ms) * time.Millisecond
+}
+
 func newRateLimitedModel() model.ToolCallingChatModel {
+	delay := getRateLimitDelay()
+	if delay == 0 {
+		return commonModel.NewChatModel()
+	}
 	return &rateLimitedModel{
 		m:     commonModel.NewChatModel(),
-		delay: time.Minute,
+		delay: delay,
 	}
 }