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

Introduction to LangGraph and Graph Theories

• The rationale for using graphs in LLM applications: comparing orchestration models versus simple linear chains.
• Understanding nodes, edges, and state within the LangGraph architecture.
• Getting started: Building and executing your first basic graph.

State Management and Prompt Chaining

• Structuring prompts as individual graph nodes.
• Managing state transitions between nodes and processing outputs effectively.
• Implementing memory strategies: Distinguishing between short-term and persistent context.

Branching, Control Flow, and Error Management

• Designing conditional routing and multi-path workflow structures.
• Implementing robust error handling, including retries, timeouts, and fallback mechanisms.
• Ensuring idempotency for safe workflow re-execution.

Tools and External Integrations

• Executing function and tool calls directly from graph nodes.
• Integrating REST APIs and external services within the graph structure.
• Handling and processing structured data outputs.

Retrieval-Augmented Workflows

• Foundations of document ingestion and text chunking.
• Utilizing embeddings and vector stores, such as ChromaDB.
• Generating grounded answers with accurate citations.

Testing, Debugging, and Quality Assurance

• Writing unit-style tests for individual nodes and workflow paths.
• Leveraging tracing and observability tools for debugging.
• Performing quality checks for factuality, safety, and determinism.

Deployment and Packaging Essentials

• Setting up environments and managing project dependencies.
• Serving graph applications via API endpoints.
• Versioning workflows and managing rolling updates.

Conclusion and Future Directions