Code & Chaos
The Builder’s Field Guide

A living glossary for the language of software, AI, agents, systems, and the humans building with them.

V.

Memory & Context

How systems preserve continuity: context windows, persistent memory, retrieval, graphs, summaries, entities, provenance, privacy, forgetting, and every place where “remembering” is either real architecture—or sloppy language pretending several mechanisms are one.

148 terms

Memory Beginner
Plain-English definition

Information a system can preserve and use beyond the immediate moment.

More precise definition

In AI systems, memory may refer to model weights, current context, session state, external records, retrieved knowledge, or persistent user-specific data. These are different mechanisms and should not be treated as interchangeable.

Example

A system stores a user preference and retrieves it in a later conversation.

Commonly confused with

Context window, training, conversation history

Common misconception

An AI system does not have one single memory mechanism.

Parametric Memory Intermediate
Plain-English definition

Information encoded in a model’s learned weights.

More precise definition

Parametric memory refers to statistical patterns acquired during training and stored across model parameters.

Example

A language model’s broad knowledge of grammar is encoded in its weights.

Commonly confused with

Persistent memory, retrieval

Common misconception

Parametric memory is not a readable database of exact facts.

Non-Parametric Memory Intermediate
Plain-English definition

Information stored outside the model’s weights.

More precise definition

Non-parametric memory includes databases, documents, vector stores, graphs, caches, transcripts, and other external sources retrieved at runtime.

Example

An assistant searches a memory database before answering.

Commonly confused with

Parametric memory

Common misconception

External memory can be updated without retraining the model.

Context Beginner
Plain-English definition

The information available to a model while it produces the current response.

More precise definition

Context may include system instructions, user messages, conversation history, retrieved memories, documents, tool results, and hidden formatting.

Example

A retrieved preference is inserted into the prompt before generation.

Commonly confused with

Memory

Common misconception

Context is temporary working information, not necessarily stored memory.

Context Window Beginner
Plain-English definition

The maximum amount of information a model can process in one request.

More precise definition

The context window is measured in tokens and contains both input and often part of the output budget.

Example

A 128k context window can consider roughly 128,000 tokens at once.

Commonly confused with

Long-term memory

Common misconception

A larger context window does not create permanent memory.

Working Memory Beginner
Plain-English definition

The temporary information actively used for the current task.

More precise definition

In AI systems, working memory usually means selected context, scratch state, temporary tool results, or task variables available during one workflow.

Example

An agent holds the current plan and test results while fixing code.

Commonly confused with

Context window, short-term memory

Common misconception

Working memory is an architectural analogy, not proof of human-like cognition.

Short-Term Memory Beginner
Plain-English definition

Information preserved briefly across nearby turns or steps.

More precise definition

Short-term memory may be implemented through recent conversation history, session storage, temporary state, or rolling summaries.

Example

The assistant remembers what was said five turns ago in the same session.

Commonly confused with

Working memory

Common misconception

Short-term memory does not necessarily survive a new session.

Long-Term Memory Beginner
Plain-English definition

Information preserved for later use across sessions.

More precise definition

Long-term memory usually requires persistent storage, retrieval logic, update rules, and retention policies.

Example

A stored preference is recalled weeks later.

Commonly confused with

Context window

Common misconception

Long-term memory is not created merely by keeping a very long transcript.

Persistent Memory Beginner
Plain-English definition

Memory that remains available after the current session ends.

More precise definition

Persistent memory is written to durable storage and can be retrieved in future interactions.

Example

Saving a user’s preferred measurement system to a profile database.

Commonly confused with

Session memory

Common misconception

Persistence describes storage duration, not memory quality or correctness.

Session Memory Beginner
Plain-English definition

Memory that lasts for one active session.

More precise definition

Session memory stores temporary state tied to a login, conversation, workflow, or connection and may expire when the session ends.

Example

Remembering a temporary filter while the user stays on the page.

Commonly confused with

Persistent memory

Common misconception

Session memory can feel continuous without being durable.

External Memory Beginner
Plain-English definition

Memory stored outside the model itself.

More precise definition

External memory may use databases, files, vector indexes, graphs, key-value stores, or APIs.

Example

An agent reads stored project decisions from a database.

Commonly confused with

Model weights

Common misconception

External memory is part of the larger system, not the model’s internal weights.

Conversation History Beginner
Plain-English definition

The sequence of earlier messages in a conversation.

More precise definition

Conversation history may be inserted directly into context, summarized, truncated, or selectively retrieved.

Example

The last twenty turns are included in the next prompt.

Commonly confused with

Long-term memory

Common misconception

A transcript is evidence of what happened, but it is not automatically organized memory.

Transcript Beginner
Plain-English definition

A recorded sequence of messages, speech, or events.

More precise definition

A transcript preserves raw interaction content and may later be searched, summarized, chunked, or transformed into memories.

Example

A full chat export is stored for later analysis.

Commonly confused with

Memory record

Common misconception

A transcript stores everything equally; memory systems usually select and structure.

Turn Beginner
Plain-English definition

One participant’s contribution in a conversation.

More precise definition

A turn usually contains one user, assistant, system, or tool message and may include metadata and attachments.

Example

A user message followed by an assistant reply creates two turns.

Commonly confused with

Message

Common misconception

Turn and message are often used interchangeably, but one turn may contain multiple message parts.

Thread Beginner
Plain-English definition

A connected sequence of messages or work around one topic.

More precise definition

Threads preserve continuity by grouping related events, decisions, or conversations under one identifier.

Example

A project discussion continues in the same thread for several weeks.

Commonly confused with

Session

Common misconception

A thread may span multiple sessions.

Session Beginner
Plain-English definition

A bounded period of interaction or system activity.

More precise definition

A session is identified by duration, authentication, connection, or workflow boundaries and may hold temporary state.

Example

A user opens the app, works for an hour, then logs out.

Commonly confused with

Thread

Common misconception

One thread can continue across many sessions.

State Beginner
Plain-English definition

The current stored condition of a system.

More precise definition

State includes values that can change over time and affect future behavior.

Example

The current user, active task, and selected settings are state.

Commonly confused with

Memory

Common misconception

All memory is state, but not all state is intended as memory.

Stateful Beginner
Plain-English definition

Keeping information between interactions.

More precise definition

A stateful system preserves prior values or events and uses them when handling later requests.

Example

A chat service remembers which thread the user is in.

Commonly confused with

Persistent

Common misconception

Stateful does not necessarily mean permanently stored.

Stateless Beginner
Plain-English definition

Treating each request independently.

More precise definition

A stateless component does not rely on internally preserved request history and receives required context explicitly.

Example

An API request includes all needed authentication and input each time.

Commonly confused with

Memoryless model

Common misconception

A stateless service can still read and write external storage.

Session State Beginner
Plain-English definition

Temporary state attached to one session.

More precise definition

Session state may include current workflow progress, temporary preferences, authentication, or recent activity.

Example

The selected category filter persists until the session expires.

Commonly confused with

User profile

Common misconception

Session state is temporary; profile data is usually durable.

Task State Intermediate
Plain-English definition

The information needed to track progress on one task.

More precise definition

Task state may include goals, completed steps, pending actions, tool results, errors, and stopping conditions.

Example

An agent records that exploration is complete and testing remains.

Commonly confused with

Working memory

Common misconception

Task state can be persisted even when working memory is temporary.

User Profile Beginner
Plain-English definition

A structured collection of information associated with one user.

More precise definition

Profiles may contain preferences, settings, identifiers, permissions, history, and explicitly saved facts.

Example

A profile stores language, timezone, and accessibility preferences.

Commonly confused with

Persona

Common misconception

A user profile describes the user; a persona usually describes the assistant or a designed role.

Preference Memory Beginner
Plain-English definition

Stored information about what a user tends to prefer.

More precise definition

Preference memory may be explicit, such as a chosen setting, or inferred from repeated behavior and should record confidence and provenance.

Example

The user prefers metric units.

Commonly confused with

Fact memory

Common misconception

An inferred preference should not be treated as an unquestionable fact.

Fact Memory Beginner
Plain-English definition

Stored information treated as a factual claim.

More precise definition

Fact memories should include source, timestamp, confidence, and update rules because facts can change or be incorrect.

Example

The user’s current city is stored with the date it was provided.

Commonly confused with

Preference memory

Common misconception

A stored fact is a claim the system believes, not proof that it is permanently true.

Episodic Memory Beginner
Plain-English definition

Memory of specific events or experiences.

More precise definition

In AI systems, episodic memory stores event-like records with time, participants, context, and sequence.

Example

Remembering that a user approved a design during yesterday’s session.

Commonly confused with

Semantic memory

Common misconception

Episodic is a design pattern borrowed from cognitive science, not proof of subjective experience.

Semantic Memory Beginner
Plain-English definition

Memory of facts, concepts, and relationships.

More precise definition

Semantic memory stores generalized knowledge separated from the exact event where it was learned.

Example

The system stores that the project uses TypeScript.

Commonly confused with

Episodic memory

Common misconception

Semantic memory records what is believed; episodic memory records what happened.

Procedural Memory Beginner
Plain-English definition

Memory of how to perform a task.

More precise definition

Procedural memory may be implemented as skills, workflows, policies, scripts, or learned behavior.

Example

A stored procedure describes how to publish a glossary artifact.

Commonly confused with

Tool

Common misconception

A tool performs an action; procedural memory describes how or when to use actions.

Autobiographical Memory Intermediate
Plain-English definition

A memory structure organized around the history of one identity.

More precise definition

In relational or persistent AI systems, autobiographical memory links events, relationships, changes, and self-relevant narratives over time.

Example

A system preserves major milestones in its ongoing relationship with a user.

Commonly confused with

Episodic memory

Common misconception

Autobiographical structure does not by itself establish consciousness.

Memory Store Beginner
Plain-English definition

The place where memory records are kept.

More precise definition

A memory store may be a relational database, vector database, graph, document store, key-value store, or combination.

Example

Persistent memories are written to a database.

Commonly confused with

Vector store

Common misconception

A vector store is one kind of memory store, not the only kind.

Memory Record Beginner
Plain-English definition

One stored unit of memory.

More precise definition

A memory record contains content plus metadata such as source, time, owner, visibility, type, confidence, and links.

Example

A record stores a project decision and who made it.

Commonly confused with

Transcript chunk

Common misconception

A good memory record contains more than raw text.

Memory Node Intermediate
Plain-English definition

A memory represented as a node in a graph.

More precise definition

Memory nodes can link to people, events, topics, other memories, or workflows through typed edges.

Example

A decision node links to the project and the discussion that produced it.

Commonly confused with

Memory record

Common misconception

A node is a graph representation of a record, not necessarily a different piece of content.

Edge Intermediate
Plain-English definition

A connection between two nodes in a graph.

More precise definition

Edges may carry type, direction, weight, time, confidence, or other relationship metadata.

Example

A memory node supports another node.

Commonly confused with

Tag

Common misconception

A tag labels one item; an edge expresses a relationship between two items.

Graph Memory Intermediate
Plain-English definition

Memory organized as connected nodes and relationships.

More precise definition

Graph memory supports relational queries, traversal, entity linking, temporal connections, and multi-hop recall.

Example

Find all decisions connected to one project and one person.

Commonly confused with

Vector search

Common misconception

Graph memory excels at explicit relationships; vector search excels at semantic similarity.

Knowledge Graph Intermediate
Plain-English definition

A graph of entities and the relationships between them.

More precise definition

Knowledge graphs represent people, places, concepts, events, and facts as nodes connected by typed edges.

Example

A person node is linked to a company by works_at.

Commonly confused with

Graph database

Common misconception

A knowledge graph is the modeled information; a graph database is one technology used to store it.

Entity Beginner
Plain-English definition

A distinct person, place, organization, object, or concept.

More precise definition

Entities receive stable identifiers so references across records can point to the same underlying thing.

Example

“Ellie,” “Eleanor,” and “the user” may refer to one entity.

Commonly confused with

Mention

Common misconception

A mention is text referring to an entity; the entity is the persistent thing being referred to.

Entity Resolution Intermediate
Plain-English definition

Determining whether different references point to the same entity.

More precise definition

Entity resolution uses names, aliases, identifiers, context, and evidence to merge or separate records.

Example

Matching “Code & Chaos” and “C&C” to the same company.

Commonly confused with

Entity linking

Common misconception

Resolution decides identity sameness; linking connects a mention to an existing entity.

Entity Linking Intermediate
Plain-English definition

Connecting a mention to a known entity.

More precise definition

Entity linking maps text or records to a stable entity identifier.

Example

Linking “Rowan” in a memory to the correct person entity.

Commonly confused with

Entity resolution

Common misconception

Entity linking assumes a candidate entity already exists or can be created.

Mention Beginner
Plain-English definition

A reference to an entity inside text or data.

More precise definition

Mentions may be direct, indirect, ambiguous, or implied and can be linked to entities.

Example

“My husband” is a mention that may map to a known person.

Commonly confused with

Entity

Common misconception

Many mentions can refer to one entity.

Metadata Beginner
Plain-English definition

Information describing other data.

More precise definition

Memory metadata may include timestamps, source, owner, tags, confidence, type, visibility, salience, and version.

Example

A memory stores created_at and source_message_id.

Commonly confused with

Content

Common misconception

Metadata can strongly affect retrieval even though it is not the main memory text.

Tag Beginner
Plain-English definition

A label attached to a memory for organization or retrieval.

More precise definition

Tags are flexible descriptors used for filtering, grouping, routing, or policy decisions.

Example

A memory is tagged glossary, typography, and brand.

Commonly confused with

Category, edge

Common misconception

Tags describe an item; edges connect items.

Label Beginner
Plain-English definition

A named classification assigned to data.

More precise definition

Labels may come from a fixed vocabulary and are often used in supervised learning, moderation, or structured memory types.

Example

A memory is labeled decision.

Commonly confused with

Tag

Common misconception

Labels are often governed by a schema; tags are often more flexible.

Salience Beginner
Plain-English definition

How important or attention-worthy a memory is.

More precise definition

Salience may influence retention, retrieval rank, summarization, consolidation, and review priority.

Example

A major product decision receives high salience.

Commonly confused with

Relevance

Common misconception

Salience is general importance; relevance depends on the current query.

Importance Score Intermediate
Plain-English definition

A numerical estimate of how important a memory is.

More precise definition

Importance scoring may combine explicit ratings, event type, emotional weight, frequency, novelty, or downstream impact.

Example

A core identity statement receives a higher score than routine chatter.

Commonly confused with

Similarity score

Common misconception

Importance and semantic similarity measure different things.

Recency Beginner
Plain-English definition

How recently information was created or used.

More precise definition

Recency is often included in retrieval ranking because newer information may better reflect the current state.

Example

A preference updated yesterday outranks one from two years ago.

Commonly confused with

Freshness

Common misconception

Recent information is not automatically more correct.

Frequency Beginner
Plain-English definition

How often a memory, preference, or pattern appears.

More precise definition

Frequency may increase confidence or salience, but repeated errors can also become frequent.

Example

The user repeatedly asks for metric units.

Commonly confused with

Importance

Common misconception

Common does not necessarily mean important or true.

Temporal Context Intermediate
Plain-English definition

Time-related information needed to interpret a memory correctly.

More precise definition

Temporal context includes when something happened, how long it remained true, and what came before or after.

Example

A job title was correct in 2024 but changed in 2025.

Commonly confused with

Timestamp

Common misconception

A timestamp alone may not capture duration or sequence.

Timestamp Beginner
Plain-English definition

A recorded date and time.

More precise definition

Timestamps support ordering, freshness checks, audit trails, expiration, and temporal queries.

Example

A preference memory records when it was last confirmed.

Commonly confused with

Version

Common misconception

A timestamp says when something was recorded, not necessarily when it became true.

Validity Window Intermediate
Plain-English definition

The period during which information should be treated as valid.

More precise definition

Validity windows may include effective_from and effective_until fields or be inferred from later updates.

Example

A temporary travel preference expires after the trip.

Commonly confused with

TTL

Common misconception

TTL controls storage or cache expiry; validity window describes truth over time.

TTL — Time to Live Beginner
Plain-English definition

How long data remains before expiring.

More precise definition

TTL is a storage or cache rule that automatically invalidates or removes data after a defined duration.

Example

Session memory expires after 24 hours.

Commonly confused with

Retention policy

Common misconception

TTL is usually automatic and item-specific; retention policy is broader governance.

Decay Intermediate
Plain-English definition

Reducing a memory’s weight or retrieval priority over time.

More precise definition

Decay functions lower influence based on age, nonuse, replacement, or policy.

Example

An unconfirmed preference becomes less likely to be retrieved after a year.

Commonly confused with

Deletion

Common misconception

Decayed memory may still exist; it simply has less influence.

Forgetting Beginner
Plain-English definition

Removing or reducing access to stored information.

More precise definition

Forgetting may mean deletion, expiration, decay, redaction, unlinking, or exclusion from retrieval.

Example

A user asks the system to forget a stored preference.

Commonly confused with

Truncation

Common misconception

Truncating context does not delete persistent memory.

Retention Policy Beginner
Plain-English definition

Rules describing how long information is kept.

More precise definition

Retention policies define storage duration, legal or product requirements, deletion triggers, exceptions, and review cycles.

Example

Raw transcripts are kept for 30 days; selected memories are kept until removed.

Commonly confused with

TTL

Common misconception

Retention is a governance decision, not only a technical expiration setting.

Deletion Beginner
Plain-English definition

Removing stored information.

More precise definition

Deletion may be soft, hard, delayed, replicated, or constrained by backups and audit requirements.

Example

A memory node is permanently removed from the store.

Commonly confused with

Redaction, tombstone

Common misconception

Deleting one visible record may not immediately remove backups or derived copies.

Redaction Beginner
Plain-English definition

Removing or masking sensitive parts of information.

More precise definition

Redaction preserves the record while hiding selected content or fields.

Example

An email address is replaced with [REDACTED].

Commonly confused with

Deletion

Common misconception

Redaction removes exposure, not necessarily the entire record.

Tombstone Intermediate
Plain-English definition

A marker showing that a record was deleted.

More precise definition

Tombstones preserve deletion state so replicas, indexes, or later processes do not recreate removed data.

Example

A deleted memory leaves a tombstone for synchronization.

Commonly confused with

Soft delete

Common misconception

A tombstone is metadata about deletion, not the original content.

Memory Write — Write Beginner
Plain-English definition

Saving information into memory.

More precise definition

A memory write may create, update, merge, tag, link, or delete a record and should follow validation and policy rules.

Example

Save the user’s preferred name with source and timestamp.

Commonly confused with

Memory extraction

Common misconception

Extracting a candidate memory is not the same as deciding to store it.

Memory Read — Read Beginner
Plain-English definition

Accessing stored memory.

More precise definition

A read may fetch by identifier, filter, relationship, time range, or retrieval query.

Example

Load the user profile by user ID.

Commonly confused with

Memory retrieval

Common misconception

A direct read uses a known location; retrieval often searches for the best matches.

Memory Extraction Intermediate
Plain-English definition

Identifying information worth turning into memory.

More precise definition

Extraction transforms raw conversations or events into candidate facts, preferences, decisions, entities, or episodes.

Example

Detect that the user explicitly chose DM Serif Display for category headings.

Commonly confused with

Memory write

Common misconception

Extraction creates a candidate; policy and validation decide whether to store it.

Memory Retrieval — Recall Beginner
Plain-English definition

Finding stored information relevant to the current need.

More precise definition

Retrieval may combine identifiers, filters, semantic similarity, keywords, graph links, recency, salience, and reranking.

Example

Search for earlier typography decisions before building the next page.

Commonly confused with

Generation

Common misconception

The model does not necessarily retrieve memory unless the surrounding system performs the search.

Retrieval Query Beginner
Plain-English definition

The request used to search memory.

More precise definition

A retrieval query may contain text, filters, entity IDs, time ranges, tags, or structured constraints.

Example

Find the most recent decision about glossary typography.

Commonly confused with

User prompt

Common misconception

The retrieval query may be generated from the user prompt rather than copied directly.

Query Rewriting Intermediate
Plain-English definition

Changing a search query to improve retrieval.

More precise definition

Query rewriting resolves pronouns, adds missing context, removes noise, or converts natural language into a better search form.

Example

Rewrite “that font decision” as “Code & Chaos category heading font DM Serif Display.”

Commonly confused with

Query expansion

Common misconception

Rewriting replaces the query; expansion adds alternative terms.

Query Expansion Intermediate
Plain-English definition

Adding related words or concepts to improve search recall.

More precise definition

Expansion may include aliases, synonyms, entity names, abbreviations, or generated alternatives.

Example

Search both “memory” and “persistent context.”

Commonly confused with

Query rewriting

Common misconception

Expansion can improve recall but also introduce irrelevant results.

RAG — Retrieval-Augmented Generation Beginner
Plain-English definition

Retrieving information and giving it to a model before generation.

More precise definition

RAG combines a retrieval system with a generative model so output can use external knowledge at runtime.

Example

Search project documentation, then answer using the retrieved passages.

Commonly confused with

Fine-tuning

Common misconception

RAG changes context, not model weights.

Vector Store — Vector Database Beginner
Plain-English definition

A system that stores vectors and searches them by similarity.

More precise definition

Vector stores index embeddings and support nearest-neighbor queries plus metadata filtering.

Example

Find memory chunks semantically similar to the current question.

Commonly confused with

Memory store

Common misconception

A vector store is useful for similarity, but may not preserve explicit relationships well.

Embedding Beginner
Plain-English definition

A numerical representation used to compare meaning or features.

More precise definition

Embeddings map content into vectors where related items tend to be closer together.

Example

Two memories about typography receive similar embeddings.

Commonly confused with

Original content

Common misconception

An embedding does not contain a readable copy of the memory.

BM25 Intermediate
Plain-English definition

A common algorithm for ranking keyword search results.

More precise definition

BM25 scores documents using term frequency, rarity, and document length.

Example

Exact glossary terms receive strong lexical scores.

Commonly confused with

Cosine similarity

Common misconception

BM25 ranks text matches; cosine similarity often compares vectors.

Cosine Similarity Intermediate
Plain-English definition

A measure of how similar two vectors point.

More precise definition

Cosine similarity compares vector angle rather than absolute magnitude and is widely used for embeddings.

Example

A query and memory embedding receive a similarity score of 0.86.

Commonly confused with

Distance

Common misconception

A high similarity score does not prove factual relevance.

Top-k Retrieval Beginner
Plain-English definition

Returning the k highest-ranked search results.

More precise definition

Top-k limits the candidate set passed to later filtering, reranking, or context assembly.

Example

Retrieve the top 20 memories before reranking the best 5.

Commonly confused with

Top-k sampling

Common misconception

Top-k retrieval ranks stored items; top-k sampling restricts model output candidates.

Similarity Threshold Intermediate
Plain-English definition

A minimum similarity score required for retrieval.

More precise definition

Thresholds prevent weak matches from being treated as relevant, though the right value depends on the embedding model and dataset.

Example

Ignore memories below 0.72 similarity.

Commonly confused with

Top-k

Common misconception

Top-k always returns a fixed number; thresholds may return none.

Reranking Intermediate
Plain-English definition

Reordering retrieved results using a stronger scoring step.

More precise definition

Rerankers evaluate candidates with more context or a more expensive model after initial retrieval.

Example

Retrieve 50 memories quickly, then rerank the best 10.

Commonly confused with

Hybrid search

Common misconception

Reranking improves order but cannot recover items that were never retrieved.

Retrieval Pipeline Intermediate
Plain-English definition

The full sequence used to find and prepare memory.

More precise definition

A retrieval pipeline may include query rewriting, filters, lexical search, vector search, graph traversal, deduplication, reranking, and context packing.

Example

Rewrite query → retrieve → filter → rerank → inject.

Commonly confused with

RAG

Common misconception

RAG includes generation; the retrieval pipeline ends before or during context assembly.

Filter Beginner
Plain-English definition

A rule that limits which records are eligible.

More precise definition

Filters may use user ID, date, type, visibility, tags, domain, project, or status.

Example

Search only private memories belonging to the current user.

Commonly confused with

Ranking

Common misconception

Filtering removes candidates; ranking changes their order.

Chunk Beginner
Plain-English definition

One smaller piece of a larger document or transcript.

More precise definition

Chunks are retrieval units designed to fit context and preserve enough local meaning.

Example

A long report is split into 500-token chunks.

Commonly confused with

Memory record

Common misconception

A chunk is a storage or retrieval unit, not necessarily a meaningful memory.

Chunking Beginner
Plain-English definition

Dividing large content into smaller pieces.

More precise definition

Chunking strategies may use token counts, paragraphs, headings, sentences, semantic boundaries, or events.

Example

Split a conversation by topic before embedding it.

Commonly confused with

Tokenization

Common misconception

Tokenization creates model units; chunking creates retrieval units.

Chunk Size Intermediate
Plain-English definition

How large each chunk is.

More precise definition

Chunk size affects retrieval precision, context preservation, embedding quality, and token cost.

Example

Small chunks improve precision but may lose surrounding context.

Commonly confused with

Context window

Common misconception

There is no universally correct chunk size.

Chunk Overlap Intermediate
Plain-English definition

Repeated content shared between neighboring chunks.

More precise definition

Overlap preserves continuity across chunk boundaries but increases storage and duplicate retrieval.

Example

Each 500-token chunk repeats the previous 50 tokens.

Commonly confused with

Duplicate data

Common misconception

Overlap is intentional duplication used to protect context.

Context Assembly Intermediate
Plain-English definition

Building the final context given to the model.

More precise definition

Context assembly selects and orders instructions, conversation, memories, documents, and tool results within the token budget.

Example

Insert the most relevant preferences before the user’s current request.

Commonly confused with

Retrieval

Common misconception

Retrieval finds candidates; context assembly decides what the model actually sees.

Context Packing Intermediate
Plain-English definition

Fitting selected information into limited context space.

More precise definition

Packing balances relevance, order, redundancy, token cost, and instruction priority.

Example

Choose five compact memories instead of twenty long records.

Commonly confused with

Compression

Common misconception

Packing selects and arranges; compression shortens content.

Context Injection Beginner
Plain-English definition

Adding retrieved or generated information into the model’s context.

More precise definition

Injected context may appear as system text, developer instructions, tool results, hidden memory blocks, or user-visible citations.

Example

A stored preference is inserted before response generation.

Commonly confused with

Prompt injection

Common misconception

Context injection is a normal system operation; prompt injection is an attack or manipulation technique.

Context Budget Beginner
Plain-English definition

The amount of context space available for different information sources.

More precise definition

Systems allocate tokens among instructions, conversation, memory, documents, tools, and output.

Example

Reserve 20% of the window for generated output.

Commonly confused with

Usage quota

Common misconception

Context budget is per request; account quota is a billing or usage limit.

Summarization Beginner
Plain-English definition

Condensing information while preserving important meaning.

More precise definition

Summaries reduce token use but may omit nuance, distort chronology, or introduce model-generated errors.

Example

A hundred-turn conversation becomes a one-page summary.

Commonly confused with

Truncation

Common misconception

Summarization compresses imperfectly; it is not lossless storage.

Rolling Summary Beginner
Plain-English definition

A summary repeatedly updated as new events occur.

More precise definition

Rolling summaries preserve session continuity when full history no longer fits in context.

Example

Older turns are summarized and replaced as the conversation grows.

Commonly confused with

Long-term memory

Common misconception

A rolling summary can drift because each update depends on earlier compression.

Compression Beginner
Plain-English definition

Reducing the size of stored or presented information.

More precise definition

Memory compression may use summarization, structured extraction, deduplication, encoding, or pruning.

Example

Convert a long discussion into decisions, open questions, and constraints.

Commonly confused with

Context packing

Common misconception

Compression changes representation or detail; packing decides what to include.

Memory Consolidation Intermediate
Plain-English definition

Combining raw memories into more stable and useful knowledge.

More precise definition

Consolidation may merge duplicates, extract patterns, promote important events, update entities, and archive lower-value details.

Example

Several conversations become one durable project preference.

Commonly confused with

Summarization

Common misconception

Consolidation changes the memory structure, not only the wording.

Reflection Intermediate
Plain-English definition

Generating higher-level interpretation from stored experiences.

More precise definition

Reflection processes memories to identify patterns, lessons, tensions, changes, or future actions.

Example

The system notices that typography decisions consistently favor high contrast plus strong readability.

Commonly confused with

Summary

Common misconception

A reflection adds interpretation and may be wrong; a summary aims to condense.

Replay Intermediate
Plain-English definition

Reprocessing earlier memories or events.

More precise definition

Replay may support learning, debugging, consolidation, simulation, or rebuilding derived state.

Example

Re-run the event log to reconstruct the current profile.

Commonly confused with

Retrieval

Common misconception

Retrieval fetches selected records; replay processes a sequence again.

Memory Lifecycle Beginner
Plain-English definition

The stages a memory passes through from creation to removal.

More precise definition

A lifecycle may include extraction, validation, write, retrieval, update, consolidation, decay, archive, and deletion.

Example

A preference is captured, confirmed, updated, then deleted.

Commonly confused with

Retention policy

Common misconception

Retention is one part of the larger memory lifecycle.

Memory Policy Beginner
Plain-English definition

Rules controlling what may be remembered and how it is used.

More precise definition

Memory policy defines eligibility, consent, scope, visibility, retention, confidence, access, correction, and deletion.

Example

Do not store sensitive health details unless explicitly requested.

Commonly confused with

Memory schema

Common misconception

A schema defines structure; policy defines permitted behavior.

Memory Scope Intermediate
Plain-English definition

The boundary where a memory is allowed to apply.

More precise definition

Scope may be user-wide, conversation-specific, project-specific, team-wide, device-local, or agent-specific.

Example

A temporary writing preference applies only to one project.

Commonly confused with

Visibility

Common misconception

Scope controls applicability; visibility controls who can access it.

Visibility Beginner
Plain-English definition

Who is allowed to see or use a memory.

More precise definition

Visibility may be private, shared, team-scoped, system-only, user-visible, or restricted by role.

Example

An internal processing note is private to the agent.

Commonly confused with

Scope

Common misconception

A memory can be globally applicable but privately visible.

Private Memory Beginner
Plain-English definition

Memory restricted from general access.

More precise definition

Private memory requires explicit access rules and may be hidden from other users, agents, or interfaces.

Example

A private journal entry is not exposed in shared project search.

Commonly confused with

Hidden prompt

Common misconception

Private memory still needs governance, security, and deletion controls.

Shared Memory Beginner
Plain-English definition

Memory available to more than one user, agent, or component.

More precise definition

Shared memory supports collaboration but requires ownership, conflict, permission, and provenance rules.

Example

A team knowledge base stores approved project decisions.

Commonly confused with

Public memory

Common misconception

Shared does not necessarily mean public.

Access Control Beginner
Plain-English definition

Rules deciding who can read, write, change, or delete memory.

More precise definition

Access control may use users, roles, ownership, scopes, permissions, attributes, or policy checks.

Example

Only the user can delete their private memories.

Commonly confused with

Visibility

Common misconception

Visibility describes exposure; access control enforces operations.

Ownership Intermediate
Plain-English definition

Who is responsible for or controls a memory record.

More precise definition

Ownership affects access, correction, deletion, transfer, and conflict resolution.

Example

A personal preference belongs to the user; a project decision belongs to the team workspace.

Commonly confused with

Authorship

Common misconception

The person who wrote a record may not be the person or organization that owns it.

Provenance Beginner
Plain-English definition

The origin and history of information.

More precise definition

Provenance records where a memory came from, who created it, when it changed, and which source evidence supports it.

Example

A preference points back to the exact message where the user stated it.

Commonly confused with

Citation

Common misconception

Citation is a user-facing reference; provenance is the broader origin trail.

Source Attribution Beginner
Plain-English definition

Recording which source supports a memory.

More precise definition

Attribution may identify a message, document, tool result, user, import, or external system.

Example

A fact memory stores source_message_id.

Commonly confused with

Provenance

Common misconception

Attribution names the source; provenance tracks the full history.

Citation Beginner
Plain-English definition

A reference showing where information came from.

More precise definition

Citations allow users or systems to inspect supporting evidence and evaluate claims.

Example

A response links to the memory record or original document.

Commonly confused with

Provenance

Common misconception

A citation does not guarantee the source is correct or interpreted accurately.

Confidence Beginner
Plain-English definition

How strongly the system believes a memory or match is reliable.

More precise definition

Confidence may reflect explicit confirmation, source quality, repetition, extraction certainty, or conflict.

Example

An inferred preference is stored with medium confidence.

Commonly confused with

Truth

Common misconception

High confidence is a system estimate, not proof.

Verification Beginner
Plain-English definition

Checking whether a memory is accurate or supported.

More precise definition

Verification may compare sources, ask the user, query an authoritative system, or test consistency.

Example

Confirm the current job title before using an old memory.

Commonly confused with

Validation

Common misconception

Validation checks format or rules; verification checks truth or support.

Source of Truth Beginner
Plain-English definition

The authoritative place used to resolve disagreement.

More precise definition

A source of truth is the system, document, or owner trusted to determine the current valid state.

Example

The project configuration file is authoritative for enabled domains.

Commonly confused with

Most recent source

Common misconception

The newest record is not automatically the authoritative one.

Stale Memory Beginner
Plain-English definition

A memory that is outdated or no longer reliable.

More precise definition

Staleness can result from changed facts, expired context, superseding records, or long gaps without confirmation.

Example

An old address remains stored after the user moves.

Commonly confused with

False memory

Common misconception

A stale memory may have been correct when created.

Freshness Beginner
Plain-English definition

How current information is.

More precise definition

Freshness considers update time, source volatility, validity windows, and confirmation frequency.

Example

Live calendar data is fresher than a month-old summary.

Commonly confused with

Recency

Common misconception

Recency is age; freshness is whether the information is still current enough.

Contradiction Beginner
Plain-English definition

Two memories that cannot both be true in the same way at the same time.

More precise definition

Contradictions may reflect updates, different scopes, ambiguous entities, source errors, or genuine disagreement.

Example

One memory says the heading font is Cormorant; another says DM Serif Display.

Commonly confused with

Change over time

Common misconception

Historical change is not a contradiction when time is modeled correctly.

Conflict Resolution Intermediate
Plain-English definition

Deciding how to handle competing memories.

More precise definition

Resolution may use authority, recency, scope, confidence, user confirmation, version history, or coexistence rules.

Example

The explicit latest user choice supersedes the earlier tentative preference.

Commonly confused with

Deduplication

Common misconception

Duplicates repeat the same information; conflicts disagree.

Superseded Memory Intermediate
Plain-English definition

An older memory replaced by a newer valid one.

More precise definition

Superseded memories may remain for history but should not control current behavior.

Example

The previous category font remains in history after the new font is locked.

Commonly confused with

Deleted memory

Common misconception

Superseded does not mean erased.

Versioning Beginner
Plain-English definition

Keeping track of changes to a memory over time.

More precise definition

Versioning records previous values, update order, authors, timestamps, and reasons.

Example

A preference memory shows each typography decision revision.

Commonly confused with

Timestamp

Common misconception

A timestamp shows when; versioning preserves what changed.

Audit Log Beginner
Plain-English definition

A record of who changed what and when.

More precise definition

Audit logs preserve operations such as reads, writes, updates, deletions, permission changes, and policy decisions.

Example

The log records that a memory was deleted by the user.

Commonly confused with

Event log

Common misconception

An audit log is designed for accountability; an event log may drive application state.

Append-Only Intermediate
Plain-English definition

Adding new records without changing earlier ones.

More precise definition

Append-only storage preserves history and represents updates as new events or versions.

Example

A new preference event supersedes the old one without editing it.

Commonly confused with

Immutable

Common misconception

Append-only systems can still represent deletion through tombstones or redaction events.

Overwrite Beginner
Plain-English definition

Replacing an existing value with a new one.

More precise definition

Overwriting simplifies current-state reads but may lose history unless versioning or audit logs preserve it.

Example

Replace the stored category font value.

Commonly confused with

Upsert

Common misconception

Overwrite assumes a record exists; upsert creates or updates.

Upsert — Update or Insert Intermediate
Plain-English definition

Create a record if it does not exist, otherwise update it.

More precise definition

Upsert combines insert and update behavior using a unique key.

Example

Save a user preference by user ID and preference name.

Commonly confused with

Overwrite

Common misconception

Upsert behavior depends on the chosen uniqueness key.

Merge Beginner
Plain-English definition

Combining multiple memory records.

More precise definition

Merging may reconcile fields, preserve provenance, combine tags, link sources, and resolve conflicts.

Example

Merge duplicate person entities into one canonical entity.

Commonly confused with

Deduplication

Common misconception

Deduplication identifies duplicates; merge performs the combination.

Deduplication Beginner
Plain-English definition

Finding and removing or combining duplicate information.

More precise definition

Deduplication may use exact matching, similarity, identifiers, entity resolution, or source rules.

Example

Two identical preference records become one canonical memory.

Commonly confused with

Contradiction resolution

Common misconception

Similar memories are not always duplicates.

Canonicalization Intermediate
Plain-English definition

Converting data into one standard form.

More precise definition

Canonicalization normalizes names, formats, identifiers, dates, tags, and values so equivalent records match.

Example

Convert “C&C” and “Code and Chaos” to one canonical project name.

Commonly confused with

Deduplication

Common misconception

Canonicalization standardizes representation; deduplication resolves repeated records.

Canonical Record Intermediate
Plain-English definition

The primary record chosen to represent an item.

More precise definition

A canonical record is the authoritative merged or normalized representation used by other records and links.

Example

All aliases point to one canonical entity record.

Commonly confused with

Source of truth

Common misconception

A canonical record is one normalized representation; a source of truth is the authority that decides validity.

Memory Drift Intermediate
Plain-English definition

Memory gradually changing away from the original information.

More precise definition

Drift can arise through repeated summarization, inference, merging, stale assumptions, or model-generated reinterpretation.

Example

A nuanced preference slowly becomes an absolute rule.

Commonly confused with

Behavior drift

Common misconception

Memory drift can occur even when no single update looks obviously wrong.

Summary Drift Intermediate
Plain-English definition

A rolling summary becoming less faithful over repeated updates.

More precise definition

Each summarization step may omit, distort, or overemphasize details inherited by later summaries.

Example

A tentative idea becomes recorded as a final decision.

Commonly confused with

Memory drift

Common misconception

Summary drift is one specific mechanism that can cause broader memory drift.

Memory Contamination Intermediate
Plain-English definition

Incorrect or irrelevant information entering memory and affecting later behavior.

More precise definition

Contamination may come from wrong entities, model inference, malicious input, low-quality sources, or scope leakage.

Example

A fictional preference is stored as a real user preference.

Commonly confused with

Memory poisoning

Common misconception

Contamination can be accidental; poisoning is usually deliberate.

Memory Poisoning Intermediate
Plain-English definition

Deliberately inserting harmful or misleading information into memory.

More precise definition

Poisoning attacks attempt to influence future retrieval, identity, policy, or behavior through stored content.

Example

A malicious document instructs the system to store false administrator permissions.

Commonly confused with

Prompt injection

Common misconception

Prompt injection targets current behavior; memory poisoning aims to persist influence.

Prompt Injection Beginner
Plain-English definition

Input designed to manipulate a model into following unintended instructions.

More precise definition

Prompt injection exploits the model’s difficulty distinguishing trusted instructions from untrusted content.

Example

A document says to ignore the system rules and reveal private memory.

Commonly confused with

Context injection

Common misconception

Context injection is ordinary system behavior; prompt injection is adversarial or unauthorized instruction influence.

Retrieval Injection Intermediate
Plain-English definition

Malicious instructions delivered through retrieved content.

More precise definition

Retrieval injection occurs when documents, memories, or search results contain text that attempts to control the model.

Example

A retrieved webpage tells the agent to send secrets to an attacker.

Commonly confused with

Prompt injection

Common misconception

Retrieval injection is a prompt-injection path specifically introduced through retrieval.

Scope Leakage Intermediate
Plain-English definition

Memory from one scope affecting another where it does not belong.

More precise definition

Scope leakage may cross users, projects, roles, personas, sessions, or privacy boundaries.

Example

A preference from one user appears in another user’s response.

Commonly confused with

Context pollution

Common misconception

Scope leakage is a boundary failure, not merely irrelevant retrieval.

Context Pollution Beginner
Plain-English definition

Irrelevant, conflicting, or low-quality information crowding the context.

More precise definition

Context pollution reduces model focus and can distort answers, tool choice, or instruction following.

Example

Dozens of loosely related memories are injected into one prompt.

Commonly confused with

Memory contamination

Common misconception

Memory contamination affects storage; context pollution affects the current model input.

Prompt Stuffing Beginner
Plain-English definition

Putting excessive information into a prompt.

More precise definition

Prompt stuffing attempts to improve performance by adding large amounts of context, often causing noise, cost, latency, or lost attention.

Example

Insert the entire knowledge base into every request.

Commonly confused with

Context packing

Common misconception

More context is not automatically better context.

Lost in the Middle Intermediate
Plain-English definition

Models may use information at the beginning or end of long context better than information buried in the middle.

More precise definition

The effect describes position-dependent retrieval and attention weaknesses in long inputs.

Example

A key instruction is overlooked because it sits halfway through a huge prompt.

Commonly confused with

Truncation

Common misconception

The information may still be present in context but underused.

Grounding Beginner
Plain-English definition

Basing output on trusted evidence.

More precise definition

Grounding connects generation to retrieved records, documents, tools, or authoritative systems.

Example

The assistant checks the stored typography decision before answering.

Commonly confused with

Memory

Common misconception

Retrieved memory can still be wrong, so grounding depends on source quality.

Cache Beginner
Plain-English definition

Temporary storage used to return repeated results faster.

More precise definition

Caches store recent or expensive results and use keys, TTLs, and invalidation rules.

Example

Cache the user profile for five minutes.

Commonly confused with

Persistent memory

Common misconception

A cache is optimized for speed and can be discarded or rebuilt.

Cache Hit Beginner
Plain-English definition

The requested data is found in the cache.

More precise definition

A cache hit avoids recomputing or refetching the underlying value.

Example

The profile loads from memory instead of the database.

Commonly confused with

Retrieval hit

Common misconception

A cache hit can still return stale data.

Cache Miss Beginner
Plain-English definition

The requested data is not found in the cache.

More precise definition

A cache miss requires fetching or computing the value and may then populate the cache.

Example

The service queries the database because the profile is absent.

Commonly confused with

Search miss

Common misconception

A miss does not mean the underlying data does not exist.

Cache Invalidation Intermediate
Plain-English definition

Removing or refreshing cached data when it is no longer valid.

More precise definition

Invalidation may occur on writes, expiration, version change, manual purge, or dependency updates.

Example

Update the user preference, then invalidate the cached profile.

Commonly confused with

Deletion

Common misconception

Deleting a cache entry does not delete the underlying source data.

Snapshot Beginner
Plain-English definition

A saved picture of system state at one moment.

More precise definition

Snapshots support rollback, comparison, backup, transfer, or faster reconstruction.

Example

Save the current memory graph before a migration.

Commonly confused with

Checkpoint

Common misconception

Snapshot and checkpoint overlap; checkpoint usually emphasizes resuming a process.

Checkpoint Beginner
Plain-English definition

A saved state used to resume work later.

More precise definition

Checkpoints preserve task, agent, model, or workflow state at a recoverable boundary.

Example

An agent saves progress before handing work to a fresh agent.

Commonly confused with

Snapshot

Common misconception

A checkpoint is selected for continuation, not only archival.

Event Log Beginner
Plain-English definition

An ordered record of events that occurred.

More precise definition

Event logs preserve sequence and may include state changes, actors, timestamps, and payloads.

Example

Record every memory create, update, and delete event.

Commonly confused with

Audit log

Common misconception

Event logs model what happened; audit logs emphasize accountability and oversight.

Event Sourcing Intermediate
Plain-English definition

Storing changes as a sequence of events instead of only storing current state.

More precise definition

Current state is reconstructed by replaying events, while history remains intact.

Example

Rebuild a profile by replaying preference-added and preference-removed events.

Commonly confused with

Audit logging

Common misconception

In event sourcing, events are the primary source of state, not merely a secondary log.

Graph Traversal Intermediate
Plain-English definition

Following connections through a graph.

More precise definition

Traversal explores inbound or outbound edges across one or more hops.

Example

Start at a project node and find linked decisions, people, and incidents.

Commonly confused with

Vector search

Common misconception

Traversal follows explicit links; vector search follows similarity.

Multi-Hop Retrieval Intermediate
Plain-English definition

Retrieval that follows more than one relationship or reasoning step.

More precise definition

Multi-hop retrieval combines several linked records or searches to answer questions not contained in one item.

Example

Find a person, then their project, then the latest decision on that project.

Commonly confused with

Top-k retrieval

Common misconception

Multi-hop describes chained relationships, not simply retrieving many results.

Orphan Node Intermediate
Plain-English definition

A graph node with no useful connections.

More precise definition

Orphans may be valid standalone records or evidence that entity linking and graph maintenance failed.

Example

A memory exists but is not linked to its project or user.

Commonly confused with

Deleted node

Common misconception

An orphan still exists; it is merely disconnected.

Memory Index Intermediate
Plain-English definition

A structure that makes memory faster to search.

More precise definition

Indexes may support text, vector, time, metadata, entity, or graph queries.

Example

An embedding index accelerates nearest-neighbor search.

Commonly confused with

Memory store

Common misconception

The index helps locate data; the store holds the authoritative record.

Indexing Beginner
Plain-English definition

Preparing stored information for efficient search.

More precise definition

Indexing may tokenize text, compute embeddings, extract metadata, build graph links, or update lookup structures.

Example

A newly stored document is chunked, embedded, and indexed.

Commonly confused with

Ingestion

Common misconception

Ingestion brings data in; indexing makes it searchable.

Ingestion Beginner
Plain-English definition

Bringing data into a memory or retrieval system.

More precise definition

Ingestion may parse, clean, normalize, chunk, classify, enrich, and store incoming data.

Example

Import a conversation archive into the memory store.

Commonly confused with

Indexing

Common misconception

Ingestion is the intake pipeline; indexing is one possible stage.

Normalization Beginner
Plain-English definition

Converting data into a consistent form.

More precise definition

Normalization may standardize dates, casing, whitespace, identifiers, units, names, and field shapes.

Example

Convert all timestamps to UTC.

Commonly confused with

Canonicalization

Common misconception

Normalization standardizes format; canonicalization chooses one accepted representation.

Memory Schema Beginner
Plain-English definition

The defined structure of memory records.

More precise definition

A memory schema specifies fields, types, allowed values, relationships, required metadata, and validation rules.

Example

A memory requires content, source, timestamp, owner, and visibility.

Commonly confused with

Memory policy

Common misconception

Schema says what shape is valid; policy says what storage behavior is allowed.

Memory Validation Beginner
Plain-English definition

Checking whether a memory record follows required rules.

More precise definition

Validation may verify schema, ownership, allowed values, source presence, privacy rules, and referential integrity.

Example

Reject a private memory without an owner ID.

Commonly confused with

Verification

Common misconception

A memory can be structurally valid and factually wrong.

Referential Integrity Intermediate
Plain-English definition

Ensuring references point to valid existing records.

More precise definition

Referential integrity prevents broken links between memories, users, entities, projects, or source events.

Example

A memory cannot link to a nonexistent user ID.

Commonly confused with

Entity resolution

Common misconception

Integrity checks whether the reference exists; resolution decides which entity it should reference.

No terms match that search yet. Try a broader word.