Zep

Building production AI agents requires precise control over what context your LLM receives. Default retrieval approaches often return too much information or miss critical details for your specific use case. This video demonstrates how Context Templates in Zep give you complete control over your agent's context block—what artifacts to retrieve, how many results to include, and how to structure the output. What You'll Learn: • How to define context templates using Zep's declarative syntax • Specifying which knowledge graph artifacts to retrieve (user summaries, facts, entities, episodes) • Filtering by custom edge types (e.g., "has_requirement", "prefers_neighborhood") • Saving and applying templates via the Zep SDK • Combining context templates with custom graph structure for domain-specific filtering • Optimizing token usage while maintaining agent reliability Key Benefits: • Focus on context engineering without managing retrieval logic • Iterate quickly to find optimal context structures • Filter to domain-specific concepts for improved reliability • Reduce token usage by retrieving only what you need Resources: Context Templates Documentation: https://help.getzep.com/v3/context-templates Customizing Graph Structure: https://help.getzep.com/customizing-graph-structure Evaluation Quickstart: https://help.getzep.com/evaluate-zep-for-your-use-case Get Started: https://app.getzep.com