Glossary

Access controls set who can read or edit files and folders in the context repository, for both workspace members and API keys.

An open specification for packaging instructions, scripts, and resources into a skill an AI agent loads on demand across different clients.

An AI agent is an AI system that uses a model to plan, choose tools, and take multi-step action toward a goal instead of returning one response.

AI skills management is the practice of creating, versioning, and controlling access to the reusable instructions a company's AI agents run.

An AI teammate is an AI presence that lives inside a company's existing chat tool, mentioned like a colleague instead of opened as a separate app.

An AI-native company builds its operations around AI agents from the start, instead of adding AI tools on top of an existing setup.

An API key is a credential that lets a tool, workflow, or agent connect to Qontext's context repository directly through its API.

Attribute-based access control grants permissions by evaluating characteristics like department, clearance, or location, instead of a single assigned role.

A business context repository is a context repository built for organizations that need to centrally maintain, version, and govern their context.

Change review is the process of approving or discarding changes before they apply to a file in the context repository.

A client is an AI tool, agent, or application connected to a workspace's context repository through an API key or via MCP.

The single, always-current place that holds everything a company knows and makes it usable to every AI agent.

Context is the information given to a model at inference time, the system prompt, retrieved documents, conversation history, and tool outputs that shape its answer.

Context agents create, edit, and synthesize files in the context repository through conversational prompts.

Context engineering is the practice of curating what tokens sit in a model's context window during a single agent run.

A context graph connects a company's entities and their relationships instead of storing knowledge as isolated documents.

A context layer is the connecting layer between a company's data and its AI agents, aggregating knowledge from fragmented sources.

Context management is the practice of organizing, updating, and governing what an AI system knows.

A context repository is the self-maintaining, versioned store of knowledge that AI agents read from.

Context rot is the gradual decline in a model's reliability on information as more tokens fill its context window.

A context window is the fixed number of tokens a model can read and write in a single request, its temporary working memory.

Continuous updates is how Qontext automatically detects and applies changes to files coming from synced data sources.

Data governance is the set of policies, roles, and controls that decide who can access, change, and be accountable for a company's data.

Data lineage tracking traces where a piece of data came from and every transformation it passed through before reaching its current form.

Converting data from different sources, formats, and conventions into one consistent format so it can be compared, merged, and reused.

An embedding is a vector, a list of numbers, that represents the meaning of a piece of text, an image, or another kind of data.

An enterprise context layer is a context layer scaled to govern knowledge across every team, tool, and agent in a large organization.

Entity resolution is the process of determining which records across different sources refer to the same thing.

A file system is how a context repository organizes company knowledge into files inside folders, each file living at one fixed address.

Fine-tuning continues training a model on a curated dataset so its weights change and its behavior shifts on every call.

Grounding is the practice of answering from real, retrieved information instead of a model's memorized guess.

Groups let an admin grant a set of people shared access to files and folders without assigning them the same workspace role.

Human in the loop means a person reviews, approves, or corrects an AI system's output at some point before it takes effect.

Hybrid search combines keyword and semantic search, merging both rankings into one result list instead of running one method alone.

An interconnected file system is a file system where files carry direct references to each other, so related knowledge stays linked.

A knowledge graph represents entities and the relationships between them, so a system can query how things connect.

Knowledge management is the practice of capturing, organizing, and sharing an organization's knowledge so people can find and reuse it.

An LLM wiki is a shared markdown knowledge base a model reads from and edits across sessions.

Member settings control workspace roles and who can invite, remove, or manage other members and shared resources.

An open standard that lets any AI application connect to external data and tools through one shared interface.

Native connectors link Qontext to external tools and keep their data continuously synced into the context repository.

An ontology defines the entity types in a domain and how they relate, the schema a knowledge graph fills in with actual records.

Open Knowledge Format packages company metadata and knowledge as markdown files with YAML frontmatter, so any AI agent can read it without a proprietary SDK.

Qontext tag is how Qontext works inside Slack: mention it in a channel or thread to get an answer from the context repository, or to have it write what the conversation just settled into the repository.

Reciprocal rank fusion merges several ranked result lists into one by scoring each result on its rank position.

RAG retrieves relevant text at query time and feeds it into a model's context window so the answer is grounded in real documents.

Retrieval logs record which context a client or file retrieval touched, so teams can trace an agent's output back to the files behind it.

Role-based access control grants permissions based on a role a person holds, rather than assigning access to each person individually.

An SDK is a packaged set of tools, libraries, and documentation a developer installs to build against a specific platform or API.

Self-attention is the mechanism that lets a model compare every token in a sequence against every other token to weigh what matters.

A semantic layer maps raw data fields to business terms like revenue or churn, so every tool and person queries the same definition.

Semantic search finds results that match a query's meaning instead of its exact words, using embeddings to compare meaning directly.

A single source of truth is the place an organization treats as authoritative for any information.

Skill management is how a company tracks the skills its AI agents use: what exists, where it lives, and who can use it.

The skill repository is where Qontext stores, versions, and distributes reusable AI instructions so every team member runs the same skill.

Source filters are the per-source settings that decide exactly which records, files, or channels a connector pulls into the context repository.

A source of truth for AI is the authoritative knowledge base an AI agent reads from, kept current and permission-aware so every agent gets the same accurate answer.

A structure file tells Qontext's agents how to organize a folder, so new content lands where your team actually wants it.

A sync is a scheduled process that pulls data from connected sources into the context repository and keeps it current.

Sync logs are the per-run record of a source's sync, showing whether a specific run succeeded, failed, or was skipped.

A transformer is the neural network architecture behind modern language models, built on self-attention.

User permissions set what workspace members can do: create, edit, and manage files, folders, sources, and access controls.

A vector database stores numeric embeddings and finds the ones closest to a query, the index behind semantic search and RAG.

Version history tracks every change to a file in the context repository, showing where each version came from.

A workspace is an isolated context repository that separates files, folders, sources, and team members so different groups can use Qontext independently.

Workspace roles are the three fixed tiers, Owner, Admin, and Member, that set a baseline of what someone can do across a Qontext workspace.