LangGraph Foundations: Graph-Based LLM Prompting and Chaining Training Course

Introduction to LangGraph and Graphical Concepts

• The rationale for utilizing graphs in LLM apps: orchestration advantages over simple chains
• Defining and understanding nodes, edges, and state within LangGraph
• Getting started: creating your first executable graph

State Management and Prompt Chaining Strategies

• Architecting prompts as discrete graph nodes
• Managing data flow by passing state between nodes and processing outputs
• Implementing memory patterns: distinguishing between short-term and persisted context

Branching Logic, Control Flow, and Error Mitigation

• Implementing conditional routing and managing multi-path workflows
• Configuring retries, timeout limits, and fallback procedures
• Ensuring idempotency and facilitating safe re-executions

Tool Integration and External Connections

• Executing function and tool calls from within graph nodes
• Interacting with REST APIs and external services inside the graph structure
• Processing and handling structured output formats

Retrieval-Augmented Generation Workflows

• Basics of document ingestion and text chunking
• Utilizing embeddings and vector stores, such as ChromaDB
• Generating grounded answers supported by citations

Testing, Debugging, and Performance Evaluation

• Writing unit-style tests for individual nodes and workflow paths
• Employing tracing and observability techniques
• Performing quality assurance checks for factuality, safety, and deterministic behavior

Packaging and Deployment Essentials

• Setting up environments and managing project dependencies
• Exposing graph workflows via API endpoints
• Managing workflow versioning and executing rolling updates

Conclusion and Future Directions