Agentic Design Patterns with LangChain

A practical learning repository for implementing agentic design patterns using LangChain framework.

Repository Structure

patterns/
├── reflection/          # ✅ Self-evaluation and improvement
├── planning/           # ✅ Task decomposition and sequential execution
├── routing/            # ✅ Query routing to specialized agents
├── parallelization/    # ✅ Concurrent task processing
├── tool_use/          # ✅ External tool integration
├── multi_agent/       # 🚧 Agent collaboration patterns
└── memory/            # 🚧 State and context management

config/               # LangChain configuration
examples/            # Complete implementations
tests/              # Pattern validation

Quick Start

1. Install dependencies: pip install -r requirements.txt
2. Set API key (optional): export OPENAI_API_KEY="your-key"
3. Run examples: python examples/reflection_example.py
4. Run tests: python -m pytest tests/

LangChain Integration

• Chains - Structured prompt → LLM → output pipelines
• Prompt Templates - Reusable, parameterized prompts
• Modular LLMs - Easy switching between providers
• Mock Support - Testing without API keys

Implemented Patterns

• ✅ Reflection Pattern - LangChain-based self-evaluation
• ✅ Prompt Chaining - Sequential prompt execution
• ✅ Routing Pattern - Query classification and specialized routing
• ✅ Parallelization Pattern - Concurrent task processing
• ✅ Tool Use Pattern - External tool integration
• 🚧 Multi-Agent Pattern - Coming soon
• 🚧 Memory Pattern - Coming soon

Learning Path

1. Start with Reflection pattern (patterns/reflection/)
2. Try Prompt Chaining (patterns/planning/)
3. Explore Routing (patterns/routing/)
4. Learn Parallelization (patterns/parallelization/)
5. Study LangChain chain composition
6. Run examples with/without API keys
7. Implement your own variations

Each pattern includes:

• LangChain implementation
• Configurable LLM support
• Usage examples
• Tests with mocks