Thank you for sending your enquiry! One of our team members will contact you shortly.
Thank you for sending your booking! One of our team members will contact you shortly.
Duration 14 hours
Course Outline
Foundations of LangGraph and Agent Patterns
- Comparing graphs vs. linear chains: applicable scenarios and rationale.
- Exploring agents, tools, and planner-executor loops.
- Building a minimal agentic graph: a practical starting point.
Managing State, Memory, and Context
- Structuring graph state and defining node interfaces.
- Differentiating between short-term and persisted memory.
- Handling context windows, summarization, and data rehydration.
Advanced Branching Logic and Control Flow
- Implementing conditional routing and multi-path decision making.
- Managing retries, timeouts, and circuit breakers.
- Designing fallbacks, handling dead-ends, and creating recovery nodes.
Integrating Tools and External Systems
- Executing function and tool calls from graph nodes and agents.
- Interacting with REST APIs and databases directly from the graph.
- Parsing and validating structured outputs.
Building Retrieval-Augmented Agent Workflows
- Strategies for document ingestion and chunking.
- Utilizing embeddings and vector stores with ChromaDB.
- Generating grounded responses with citations and safety guards.
Evaluation, Debugging, and Observability
- Tracing execution paths and analyzing node interactions.
- Implementing golden sets, evaluations, and regression tests.
- Monitoring quality, safety, and cost versus latency metrics.
Deployment and Packaging
- Serving applications via FastAPI and managing dependencies.
- Versioning graphs and defining rollback strategies.
- Establishing operational playbooks and incident response protocols.
Wrap-up and Future Directions
Requirements
- Solid proficiency in Python.
- Prior experience in developing LLM applications or managing prompt chains.
- Understanding of REST APIs and JSON data structures.
Target Audience
- AI Engineers.
- Product Managers.
- Developers focused on creating interactive, LLM-powered systems.