Skip to content

samples

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

samples

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
csharp
csharp
 
 
go
go
 
 
java
java
 
 
python
python
 
 
rust
rust
 
 
typescript
typescript
 
 
README.md
README.md
 
 

README.md

Samples

Runnable code samples for all 22 Agent Integration Patterns.

Structure

samples/
  python/               # Python samples — LangChain / LangGraph
    messaging/ discovery/ context/ routing/ coordination/ resilience/ security/ evaluation/
    requirements.txt
  java/                 # Java samples — Spring AI 1.0.0 + Mockito tests
    pom.xml
    src/main/java/com/agents/patterns/<category>/
    src/test/java/com/agents/patterns/<category>/
  typescript/           # TypeScript samples — LangChain.js / LangGraph.js
    messaging/ discovery/ context/ routing/ coordination/ resilience/ security/ evaluation/
    package.json  tsconfig.json
  csharp/               # C# samples — Microsoft Semantic Kernel
    src/AgentPatterns/<category>/
    tests/AgentPatterns.Tests/<category>/
    AgentPatterns.sln

Python

Requirements: Python 3.10+, ANTHROPIC_API_KEY set.

cd samples/python
pip install -r requirements.txt

# Run tests (no API key needed — uses mocks)
python3 -m pytest tests/ -v    # 63 tests, all patterns covered

# Run samples against real API
export ANTHROPIC_API_KEY=your-key-here
python3 messaging/direct_message.py
python3 routing/pipeline.py
python3 evaluation/ensemble_judge.py

All samples use claude-haiku-4-5-20251001 for fast, low-cost execution.

Category File Pattern
messaging direct_message.py Direct Message
messaging broadcast_message.py Broadcast Message
messaging blackboard.py Blackboard
discovery agent_card_registry.py Agent Card Registry
discovery agent_proxy.py Agent Proxy
context context_injection.py Context Injection
context tool_provider.py Tool Provider
routing content_based_router.py Content-Based Router
routing scatter_gather.py Scatter-Gather
routing pipeline.py Pipeline
coordination supervised_delegation.py Supervised Delegation
coordination orchestrator.py Orchestrator
coordination choreography.py Choreography
coordination peer_to_peer_delegation.py Peer-to-Peer Delegation
resilience dead_letter_agent.py Dead Letter Agent
resilience circuit_breaker.py Circuit Breaker
resilience checkpoint_resume.py Checkpoint & Resume
security least_privilege_tool_scope.py Least-Privilege Tool Scope
security trust_boundary.py Trust Boundary
security prompt_firewall.py Prompt Firewall
evaluation llm_as_judge.py LLM-as-Judge
evaluation ensemble_judge.py Ensemble Judge

Java (Spring AI)

Requirements: Java 21+, Maven 3.6+. Tests use Mockito — no API key needed.

cd samples/java
mvn test        # compile + run all 32 Mockito tests (no API key required)

To run against a real Anthropic API, set the key in application.properties:

spring.ai.anthropic.api-key=your-key-here
Category Implementation Test
messaging DirectMessage.java DirectMessageTest.java
messaging BroadcastMessage.java BroadcastMessageTest.java
messaging Blackboard.java BlackboardTest.java
discovery AgentCardRegistry.java AgentCardRegistryTest.java
discovery AgentProxy.java AgentProxyTest.java
context ContextInjection.java ContextInjectionTest.java
context ToolProvider.java ToolProviderTest.java
routing ContentBasedRouter.java ContentBasedRouterTest.java
routing ScatterGather.java ScatterGatherTest.java
routing Pipeline.java PipelineTest.java
coordination SupervisedDelegation.java SupervisedDelegationTest.java
coordination Orchestrator.java OrchestratorTest.java
coordination Choreography.java ChoreographyTest.java
coordination PeerToPeerDelegation.java PeerToPeerDelegationTest.java
resilience DeadLetterAgent.java DeadLetterAgentTest.java
resilience CircuitBreaker.java CircuitBreakerTest.java
resilience CheckpointResume.java CheckpointResumeTest.java
security LeastPrivilegeToolScope.java LeastPrivilegeToolScopeTest.java
security TrustBoundary.java TrustBoundaryTest.java
security PromptFirewall.java PromptFirewallTest.java
evaluation LlmAsJudge.java LlmAsJudgeTest.java
evaluation EnsembleJudge.java EnsembleJudgeTest.java

TypeScript (LangChain.js / LangGraph.js)

Requirements: Node.js 18+, ANTHROPIC_API_KEY set.

cd samples/typescript
npm install
export ANTHROPIC_API_KEY=your-key-here

# Run any pattern
npx tsx messaging/direct_message.ts
npx tsx routing/pipeline.ts
npx tsx evaluation/ensemble_judge.ts

# Type-check all 22 files (no API key needed)
npx tsc --noEmit

All samples use claude-haiku-4-5-20251001.

Category File Pattern
messaging direct_message.ts Direct Message
messaging broadcast_message.ts Broadcast Message
messaging blackboard.ts Blackboard
discovery agent_card_registry.ts Agent Card Registry
discovery agent_proxy.ts Agent Proxy
context context_injection.ts Context Injection
context tool_provider.ts Tool Provider
routing content_based_router.ts Content-Based Router
routing scatter_gather.ts Scatter-Gather
routing pipeline.ts Pipeline
coordination supervised_delegation.ts Supervised Delegation
coordination orchestrator.ts Orchestrator
coordination choreography.ts Choreography
coordination peer_to_peer_delegation.ts Peer-to-Peer Delegation
resilience dead_letter_agent.ts Dead Letter Agent
resilience circuit_breaker.ts Circuit Breaker
resilience checkpoint_resume.ts Checkpoint & Resume
security least_privilege_tool_scope.ts Least-Privilege Tool Scope
security trust_boundary.ts Trust Boundary
security prompt_firewall.ts Prompt Firewall
evaluation llm_as_judge.ts LLM-as-Judge
evaluation ensemble_judge.ts Ensemble Judge

C# (Microsoft Semantic Kernel)

Requirements: .NET 9 SDK. Tests use Moq — no API key needed.

cd samples/csharp
dotnet test        # compile + run all 34 tests (no API key required)

To run a specific pattern against a real Anthropic API, set the key in your environment:

export ANTHROPIC_API_KEY=your-key-here
Category Implementation Test
messaging Messaging/DirectMessage.cs Messaging/DirectMessageTests.cs
messaging Messaging/BroadcastMessage.cs Messaging/BroadcastMessageTests.cs
messaging Messaging/Blackboard.cs Messaging/BlackboardTests.cs
discovery Discovery/AgentCardRegistry.cs Discovery/AgentCardRegistryTests.cs
discovery Discovery/AgentProxy.cs Discovery/AgentProxyTests.cs
context Context/ContextInjection.cs Context/ContextInjectionTests.cs
context Context/ToolProvider.cs Context/ToolProviderTests.cs
routing Routing/ContentBasedRouter.cs Routing/ContentBasedRouterTests.cs
routing Routing/ScatterGather.cs Routing/ScatterGatherTests.cs
routing Routing/Pipeline.cs Routing/PipelineTests.cs
coordination Coordination/SupervisedDelegation.cs Coordination/SupervisedDelegationTests.cs
coordination Coordination/Orchestrator.cs Coordination/OrchestratorTests.cs
coordination Coordination/Choreography.cs Coordination/ChoreographyTests.cs
coordination Coordination/PeerToPeerDelegation.cs Coordination/PeerToPeerDelegationTests.cs
resilience Resilience/DeadLetterAgent.cs Resilience/DeadLetterAgentTests.cs
resilience Resilience/CircuitBreaker.cs Resilience/CircuitBreakerTests.cs
resilience Resilience/CheckpointResume.cs Resilience/CheckpointResumeTests.cs
security Security/LeastPrivilegeToolScope.cs Security/LeastPrivilegeToolScopeTests.cs
security Security/TrustBoundary.cs Security/TrustBoundaryTests.cs
security Security/PromptFirewall.cs Security/PromptFirewallTests.cs
evaluation Evaluation/LlmAsJudge.cs Evaluation/LlmAsJudgeTests.cs
evaluation Evaluation/EnsembleJudge.cs Evaluation/EnsembleJudgeTests.cs

Rust

Requirements: Rust 1.75+ (2021 edition), Cargo. No API key needed — all examples use a deterministic FakeAgent.

cd samples/rust

# Check all examples compile
cargo check

# Run any pattern
cargo run --example messaging_direct_message
cargo run --example routing_pipeline
cargo run --example resilience_circuit_breaker
cargo run --example evaluation_reflection_loop
# etc.

All examples run fully offline. Each is a self-contained examples/<category>_<pattern>.rs file using FakeAgent from src/lib.rs.

Category Example Pattern
messaging messaging_direct_message Direct Message
messaging messaging_broadcast_message Broadcast Message
messaging messaging_blackboard Blackboard
discovery discovery_agent_card_registry Agent Card Registry
discovery discovery_agent_proxy Agent Proxy
discovery discovery_broker Broker
context context_context_injection Context Injection
context context_tool_provider Tool Provider
routing routing_content_based_router Content-Based Router
routing routing_scatter_gather Scatter-Gather
routing routing_pipeline Pipeline
coordination coordination_orchestrator Orchestrator
coordination coordination_supervised_delegation Supervised Delegation
coordination coordination_choreography Choreography
coordination coordination_peer_to_peer_delegation Peer-to-Peer Delegation
coordination coordination_mediator Mediator Agent
coordination coordination_group_chat Group Chat
coordination coordination_saga Saga / Compensating Action
coordination coordination_magentic Magentic
resilience resilience_circuit_breaker Circuit Breaker
resilience resilience_dead_letter_agent Dead Letter Agent
resilience resilience_checkpoint_resume Checkpoint & Resume
implementation implementation_idempotent_agent Idempotent Agent
security security_least_privilege_tool_scope Least-Privilege Tool Scope
security security_trust_boundary Trust Boundary
security security_prompt_firewall Prompt Firewall
evaluation evaluation_llm_as_judge LLM-as-Judge
evaluation evaluation_ensemble_judge Ensemble Judge
evaluation evaluation_reflection_loop Reflection Loop

Go

Requirements: Go 1.22+. No API key needed — all patterns use a deterministic FakeAgent from shared/fake_agent.go. No external dependencies (stdlib only).

cd samples/go

# Run all patterns at once
go run ./cmd/main.go

# Run a single pattern file directly
go run ./messaging/direct_message.go ./shared/fake_agent.go

# Build check
go build ./...

Each pattern file exposes a Run() function. cmd/main.go calls all 29 of them in sequence.

Category File Pattern
messaging messaging/direct_message.go Direct Message
messaging messaging/broadcast_message.go Broadcast Message
messaging messaging/blackboard.go Blackboard
discovery discovery/agent_card_registry.go Agent Card Registry
discovery discovery/agent_proxy.go Agent Proxy
discovery discovery/broker.go Broker
context context/context_injection.go Context Injection
context context/tool_provider.go Tool Provider
routing routing/content_based_router.go Content-Based Router
routing routing/scatter_gather.go Scatter-Gather
routing routing/pipeline.go Pipeline
coordination coordination/orchestrator.go Orchestrator
coordination coordination/supervised_delegation.go Supervised Delegation
coordination coordination/choreography.go Choreography
coordination coordination/peer_to_peer_delegation.go Peer-to-Peer Delegation
coordination coordination/mediator.go Mediator Agent
coordination coordination/group_chat.go Group Chat
coordination coordination/saga.go Saga / Compensating Action
coordination coordination/magentic.go Magentic
resilience resilience/circuit_breaker.go Circuit Breaker
resilience resilience/dead_letter_agent.go Dead Letter Agent
resilience resilience/checkpoint_resume.go Checkpoint & Resume
implementation implementation/idempotent_agent.go Idempotent Agent
security security/least_privilege_tool_scope.go Least-Privilege Tool Scope
security security/trust_boundary.go Trust Boundary
security security/prompt_firewall.go Prompt Firewall
evaluation evaluation/llm_as_judge.go LLM-as-Judge
evaluation evaluation/ensemble_judge.go Ensemble Judge
evaluation evaluation/reflection_loop.go Reflection Loop