Long-Term Memory

Short-term memory is ephemeral — it captures what happened in this conversation, but not what the agent knows about the world. This module covers the persistent knowledge graphiFull definition for knowledge graph (opens in a new tab)A representation of real-world entities and their relationships, stored according to organizing principles, typically in a graph database. layer: how to classify and store entities using the POLE+O model, why a graph outperforms a vector database for multi-hop knowledge retrieval, and how the entity extraction pipeline automatically populates this layer from conversation messages.

By the end of this module, you will:

• Classify any business domain using the POLE+O entity model
• Explain why multi-hop graph traversaliFull definition for traversal (opens in a new tab)Following relationships from one node to the next to reach other parts of a graph. answers questions that vector databases cannot
• Configure the three-stage entity extraction pipeline (spaCy → GLiNER2 → LLMiGo to glossary for large language model (opens in a new tab)A model trained on text to predict the next token, and so to generate language. fallback)
• Store, search, and retrieve entities using the long-term memory API
• Run the extraction pipeline against your own instance and follow EXTRACTED_FROM back to the message an entity came from