Glossary category
Context and knowledge
How systems select, preserve, retrieve, and structure the information a model can use.
What belongs here
Context windows, retrieval, memory, embeddings, knowledge graphs, and grounding.
What belongs elsewhere
General model training and workflow control that does not concern information state.
Adjacent categories
Start with
Factory areas
Essays
Definitions in this category
stable
Knowledge graph
A structured representation of entities, concepts, and claims connected by named relationships and provenance.
Reranking
Rescoring an initial set of retrieved candidates with a more precise model or rule before selecting context.
Retrieval-augmented generation (RAG)
Generating a response after retrieving relevant material from an external knowledge source and adding it to model context.
Semantic search
Retrieval based primarily on meaning represented by embeddings rather than exact keyword overlap.
Vector database
A data system designed to store embeddings and retrieve items by vector similarity, often with metadata filtering.
working
Agent memory
State preserved outside a single model call and made available to influence later agent decisions.
Chunking
Splitting documents or data into retrieval units that can be indexed, selected, and placed into model context.
Context rot
The degradation in an agent's ability to find, follow, or correctly weigh relevant information as its accumulated context becomes longer, noisier, stale, repetitive, or internally conflicting.
Durable memory
State intentionally retained across runs, such as verified facts, decisions, preferences, learned procedures, or persistent identity.
GraphRAG
Retrieval-augmented generation that builds and queries graph structure, often alongside source text and vector retrieval.
Prompt caching
Reusing computation for repeated prompt prefixes or context blocks to reduce inference latency and cost.
Prompt compression
Reducing the tokens sent to a model while attempting to preserve the instructions, evidence, state, and relationships needed for the task.
Working memory
Short-lived task state actively used during a run, such as the current plan, recent observations, pending actions, and temporary summaries.