> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.apologist.ai/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.apologist.ai/_mcp/server.

# Glossary of Terms

> Definitions for Apologist Console, Agent, evaluator, content, and API concepts

Use this glossary to understand terminology shared across Apologist AI Console, Agents, evaluators, content workflows, and the Agent API.

[A](#a) · [B](#b) · [C](#c) · [D](#d) · [E](#e) · [F](#f) · [G](#g) · [H](#h) · [I](#i) · [J](#j) · [K](#k) · [L](#l) · [M](#m) · [N](#n) · [O](#o) · [P](#p) · [R](#r) · [S](#s) · [T](#t) · [U](#u) · [V](#v) · [W](#w) · [Y](#y)

## A

**Agent**

An Apologist conversational AI configured with instructions, sources, branding, and behavior settings to answer questions in a values-aligned way.

**Agent Sync**

A process that keeps an Agent's configuration and related content up to date across environments or connected surfaces.

**Always**

A Timing Mode that allows the action to run whenever its other conditions match.

**Anthology**

A Collection Type for a curated set of works gathered under a shared theme or purpose.

**AO Lab**

A Bible Translation Provider associated with AO Lab scripture data used by Agents.

**API (Application Programming Interface)**

A programmatic interface that lets applications request Agent responses and related Apologist services over HTTP.

**API (Client Type)**

A Client Type indicating the request came from a direct API integration.

**API Key**

A secret credential used to authenticate requests to the Apologist Agent API.

**API.Bible**

A Bible Translation Provider that delivers scripture text through the API.Bible platform.

**Article**

A Source Type for article- or paper-length written content.

**Attribution**

Credit given to a source, contributor, or organization when content is cited or displayed in a response.

**Audience**

A defined group of users used by evaluators and automations to target matching, routing, or follow-up behavior.

**Author**

A Contributor Role for the primary writer of a work.

**Auto Recharge**

A billing setting that automatically purchases additional credits when a team's balance falls below a threshold.

**Automation**

An evaluator-driven workflow that triggers actions—such as CTAs, routing, or notifications—based on conversation signals.

## B

**Behavior**

The configurable response style and constraints that shape how an Agent answers, including tone, guardrails, and output preferences.

**Benchmark**

A structured evaluation run that scores an Agent against a fixed set of questions to measure quality over time.

**Benchmark (Client Type)**

A Client Type indicating the request was produced by a benchmark run.

**Benchmark Question**

A single prompt used in a benchmark to test how an Agent responds under controlled conditions.

**Benchmark Question Variant**

An alternate wording of a benchmark question used to test consistency across paraphrases.

**Bible Translation**

A specific Bible text version an Agent can quote or reference when generating responses.

**Bible Translation Provider**

A service or publisher that supplies Bible translation text and metadata for use by Agents.

**Bibliography**

A formatted list of sources cited or consulted in connection with an Agent response or content item.

**Book**

A Source Type for book-length works ingested into the corpus.

**Boost \[Source]**

A ranking adjustment that increases a source's likelihood of being retrieved and cited relative to peers.

**Branding**

Visual and textual customization—such as colors, logos, and chrome—applied to an Agent's interface.

**Bulk Import**

An operation that adds many sources or listings to the platform in a single batch.

**Bury \[Source]**

A ranking adjustment that decreases a source's likelihood of being retrieved and cited relative to peers.

## C

**Cache**

Temporarily stored data or responses reused to improve performance and reduce repeated computation.

**Category**

A taxonomic label used to organize sources, listings, or other content for discovery and filtering.

**Channel**

A messaging or delivery surface—such as Discord, SMS, or WhatsApp—through which users interact with an Agent.

**Channel (Client Type)**

A Client Type indicating the request came through a messaging Channel.

**Chat Logs**

Stored records of Agent conversations used for review, analytics, evaluation, and debugging.

**Citation**

A reference in an Agent response that points back to a specific source used as evidence.

**Classifier**

A model or rule component that labels conversation content—such as intent, topic, or risk—for evaluators and routing.

**Client Type**

The kind of client surface initiating an Agent request, which can affect available features and telemetry.

**Collection**

A curated grouping of sources managed together for retrieval, publishing, or Agent configuration.

**Collection Type**

The structural kind of a Collection, describing how its member sources are organized.

**Commentator**

A Contributor Role for someone providing commentary on a text or topic.

**Community Sources**

Sources contributed or shared at the broader Apologist community level rather than owned solely by one team.

**Completion**

An Agent response generated for a prompt, typically corresponding to a chat completion API result.

**Confidence Threshold**

A numeric cutoff that determines whether a score or classifier result is high enough to trigger an action.

**Contributor**

A person credited on sources or listings, such as an author, speaker, or host.

**Contributor Role**

The capacity in which a Contributor participated in a source, such as author or translator.

**Conversation**

A multi-turn exchange between a user and an Agent within a session.

**Corpus**

The searchable body of indexed sources available for retrieval and citation by Agents.

**Credit**

A unit of prepaid usage consumed when Agents generate responses or run billable features.

**Credit Bundle**

A packaged purchase of Credits sold as a fixed quantity for team usage.

**CTA (Call to Action)**

A prompted next step shown to users—such as a link, offer, or follow-up—often driven by evaluators.

## D

**Data**

A Source Type for structured datasets rather than narrative documents.

**Dataset**

A Collection Type for structured or bulk data treated as a coherent source set.

**Denominational Alignment**

A classification describing how content or an Agent relates to particular Christian traditions.

**Device**

The end-user hardware or client environment associated with a session or analytics event.

**Disclaimer**

Notice text shown with an Agent or response to set expectations or legal boundaries.

**Discord**

A Channel integration that connects an Agent to Discord servers and interactions.

## E

**Editor**

A Contributor Role for someone who prepared or revised a work for publication.

**Embed**

An Agent interface placed inside a host webpage so visitors can chat without leaving the site.

**Embedded**

A Client Type for Agents embedded in a host website or application UI.

**ESV**

The English Standard Version Bible translation available for Agent scripture references.

**Evaluator**

A configured AI or rules system that inspects conversations and can score, route, flag, or trigger automations.

**External ID**

An identifier from an outside system used to correlate Apologist records with your own data.

## F

**Favicon**

The small icon representing an Agent or site in browser tabs and bookmarks.

**Feedback**

User-submitted ratings or comments about an Agent response, used for quality review.

**Flagged**

A state indicating a conversation, user, or response has been marked for review or special handling.

**Footer**

Persistent end matter on an Agent interface or response, often used for disclaimers or links.

**Foreword**

A Contributor Role for someone who wrote introductory remarks for a work.

**Frequency Penalty**

A model parameter that reduces repetition by penalizing tokens that have already appeared frequently.

**Full Bleed Media**

Media displayed edge-to-edge in the Agent UI without surrounding padding or card chrome.

## G

**Git Sync**

A sync method that updates content or configuration from a Git repository.

**Grader**

A component that assigns scores to benchmark answers or evaluator outcomes.

**Guardrail**

A constraint that blocks, rewrites, or steers unsafe or out-of-policy Agent behavior.

## H

**Hierarchy**

A parent/child organization of content—such as collections and sources—used for inheritance and navigation.

**History**

Prior turns in a conversation that the Agent can use as context for the next response.

**Host**

A Contributor Role for the host of a podcast, program, or conversation.

## I

**Import**

The process of bringing external content into Apologist as sources or listings.

**Impression**

An analytics event recorded when a corpus item, CTA, or other element is shown to a user.

**Index**

The searchable representation of source content used by semantic retrieval.

**Inherit**

A setting behavior where a child resource receives configuration values from its parent.

**Input (Token)**

Tokens in the prompt and context sent into the model.

**Instructions**

Natural-language guidance that steers an Agent's persona, priorities, and response policies.

**Interval (Timing Mode)**

A Timing Mode that allows the action only after a configured time or turn gap.

**Interviewee**

A Contributor Role for someone interviewed as part of a source.

**Intro**

Introductory content shown at the start of an Agent experience before the conversation begins.

## J

**JSON**

A structured data format used for API payloads and machine-readable Agent outputs.

## K

**Knowledge Base**

A practical term for the corpus and collections an Agent can retrieve from when answering.

## L

**Learn More URL**

A link associated with a source or listing where users can explore the full resource.

**Liked**

A user signal indicating a positive reaction to a specific Agent completion.

**List (Collection Type)**

A Collection Type representing an ordered or unordered list of related sources.

**Listing**

A public or reviewable profile for a contributor, organization, or resource on Apologist Social.

**Live Web Search**

A capability that lets an Agent query the live web for fresh information beyond the indexed corpus.

**LLM (Large Language Model)**

The underlying AI model that generates Agent text from prompts, instructions, and retrieved context.

**Lock Expiry**

The time when a lock automatically ends and editing becomes available again.

**Locked**

A state where a resource or setting cannot be edited until it is unlocked or a lock expires.

## M

**Markdown**

A lightweight markup format commonly used for Agent responses and documentation content.

**Match**

A Ranking Factor Mode that emphasizes how well a candidate matches the query or criteria.

**Matching Options**

Configuration that controls how evaluators or CTAs decide whether a conversation matches criteria.

**Max Output Tokens**

A limit on how many tokens an Agent may generate in a single response.

**MCP (Client Type)**

A Client Type for clients speaking the Model Context Protocol.

**MCP (Model Context Protocol)**

An open protocol for connecting AI clients to tools and context providers, including Apologist Agents.

**Media File**

A Source Type for uploaded media assets such as audio or document files.

**Memory**

Information retained across turns or sessions so an Agent can stay consistent with earlier context.

**Merge Tag**

A placeholder token replaced at runtime with user, audience, or conversation values.

**Messenger**

A Channel integration that connects an Agent to Facebook Messenger conversations.

**Metadata**

Structured descriptive fields attached to a source or object, such as titles, dates, and identifiers.

**Modal**

A dialog overlay in the Agent UI used for focused prompts, confirmations, or secondary flows.

**Model**

A specific LLM or voice model version selected for generating Agent output.

**Model Class**

A grouping of related models that share capability tier or pricing characteristics.

**Model Provider**

The vendor or platform that hosts a given LLM or voice model.

**Multiply**

A Ranking Factor Mode that scales scores by multiplying contributing factors.

## N

**Notification**

An alert delivered when an event occurs, such as a webhook, email, or in-product notice.

**Numeric Value**

A quantitative field extracted or assigned during evaluation for threshold comparisons.

**Numeric Value Threshold**

The cutoff applied to a numeric value when deciding whether an evaluator action should fire.

## O

**Once**

A Timing Mode that allows the action to run only a single time per conversation or scope.

**Open Source**

Content or software made available under terms that allow public use, study, or redistribution.

**OpenAI Compatible Stream**

A streaming response format that follows OpenAI-style chat completion chunk events.

**Organization**

An entity listing representing a ministry, publisher, or other institution in the Apologist ecosystem.

**Output (Token)**

Tokens generated by the model in the completion.

## P

**Password Protection**

Access control that requires a password before users can open an Agent or protected surface.

**Periodical**

A Collection Type for serial publications such as journals or magazines.

**Persisted Responder**

A responder assignment that remains applied across turns until it is changed or cleared.

**Platform**

The Apologist product environment where teams manage Agents, content, billing, and analytics.

**Platform Fee**

A charge applied by Apologist for usage or subscription services beyond base model costs.

**Podcast (Collection Type)**

A Collection Type representing a podcast feed and its episodes.

**Podcast Episode**

A Source Type for an individual podcast episode.

**Presence Penalty**

A model parameter that encourages new topics by penalizing tokens already present in the text.

**Producer**

A Contributor Role for someone who produced media content.

**Prompt**

The user or system input that initiates or continues an Agent completion.

**Prompt (Token)**

Tokens attributable to prompt content within billed or reported usage.

**PWA (Progressive Web App)**

An installable web Agent experience that can behave like a native app on supported devices.

## R

**Randomize**

A Ranking Factor Mode that introduces controlled randomness into candidate ordering.

**Ranking Factor**

A signal used when ordering retrieved sources or candidates for inclusion in a response.

**Ranking Factor Mode**

The method used to apply a ranking factor when combining or adjusting scores.

**Raw Score**

The unscored or pre-normalized numeric result produced before thresholds or ranking adjustments.

**Realtime Translation**

On-the-fly translation of user input or Agent output across languages during a conversation.

**Reasoning (Token)**

Tokens used for internal model reasoning when a provider exposes that usage.

**Reasoning Level**

A control for how much deliberate chain-of-thought style computation a model should apply.

**Referral**

An event or link that sends a user from an Agent citation or CTA to an external destination.

**Related Media**

Supplementary images, audio, or video shown alongside Agent responses or the Agent UI.

**Relevance**

A measure of how well retrieved content matches the user's question or conversation context.

**Relevance Threshold**

The minimum relevance score required for a source to be used or shown.

**Responder**

A specialized response path or persona an evaluator can select for handling a conversation.

**Responder Routing**

Evaluator logic that chooses which responder should handle the next part of a conversation.

**Response**

The content returned by an Agent for a given turn, including text and any attached metadata.

**Response Basis**

An explanation of what grounded a response, such as retrieved sources or model knowledge.

**Response Footer**

Secondary text appended after the main response body, often for citations or notices.

**Response Interval**

A timing setting that controls how often a repeated or sequenced response may appear.

**Response Sequence**

An ordered set of evaluator responses or messages delivered across matching turns.

**RSS Feed**

A Collection Type backed by an RSS or Atom feed of periodically updated items.

## S

**Score**

A numeric assessment produced by a grader, evaluator, or ranking system.

**Score Threshold**

The minimum score required for a result to count as a pass or to trigger an action.

**Selector**

A UI or configuration control used to choose among options such as models, sources, or audiences.

**Semantic Search**

Retrieval that finds corpus items by meaning rather than exact keyword overlap.

**Series**

A Collection Type for sequenced content meant to be consumed in order.

**Session**

A contiguous period of Agent interaction that groups messages, state, and analytics for a user.

**Set Responder Link**

A link or action that assigns a specific responder for subsequent conversation turns.

**Set User Flag Link**

A link or action that applies a user flag when visited or invoked.

**Share Token**

A secret identifier that grants access to a shared conversation view.

**Share URL**

A public or restricted link used to open a shared conversation.

**Slug**

A URL-friendly identifier for a page, Agent, or content object.

**SMS**

A Channel integration that lets users message an Agent over text messaging.

**Source**

A unit of content in the corpus—such as a book, article, or web page—that Agents can retrieve and cite.

**Source Persistence**

Settings that control whether and how a source remains available or pinned across contexts.

**Source Text**

The body of a source used for indexing and retrieval, distinct from its metadata.

**Source Type**

The format category of a source, which affects ingestion and display.

**Source Version**

A specific revision of a source's content used for indexing and citation stability.

**Source Weight**

A relative importance value that influences how strongly a source competes in retrieval.

**Speaker**

A Contributor Role for someone who delivered spoken content such as a talk or sermon.

**Standalone**

A Client Type for the full Agent experience running as its own app or site.

**Starter Prompt**

A suggested question or prompt shown to help users begin a conversation with an Agent.

**Streaming Video**

A Source Type for video content hosted on streaming platforms.

**Subscription**

A recurring plan that entitles a team to platform features and usage allowances.

**System Prompt**

Hidden model-level instructions that establish core Agent behavior behind the scenes.

## T

**Tag**

A label applied to sources, prompts, or other objects for filtering and organization.

**Team**

The workspace account that owns Agents, content, members, billing, and settings in Console.

**Team Sources**

Sources owned and managed by a specific team rather than the broader community corpus.

**Telegram**

A Channel integration that connects an Agent to Telegram chats and bots.

**Temperature**

A model parameter controlling randomness; higher values produce more varied outputs.

**Threshold (Timing Mode)**

A Timing Mode that fires when a score or metric crosses a configured threshold.

**Timing Mode**

A setting that determines when an evaluator action or sequenced response is allowed to fire.

**Token**

A unit of text processed by an LLM, used for usage metering and context limits.

**Top P**

A nucleus-sampling parameter that limits generation to a probability mass of candidate tokens.

**Tracking**

Measurement of user and content events for analytics, attribution, and optimization.

**Translator**

A Contributor Role for someone who translated a work into another language.

**Trial**

A time- or usage-limited period for evaluating Apologist features before full commitment.

**Twilio**

A messaging provider used to power SMS and WhatsApp Channel deliveries for Agents.

## U

**Usage**

Metered consumption of credits, tokens, or features by a team or Agent.

**User**

An end person interacting with an Agent or represented in analytics and flags.

**User Flag**

A label attached to a user that evaluators and automations can read or set.

## V

**Vanity Domain**

A custom domain used to host an Agent experience under your own brand.

**Verbosity**

A control for how concise or expansive an Agent's responses should be.

**View**

An analytics event recorded when a user opens or focuses a corpus item or page.

**Voice Model**

A speech model used to synthesize or process spoken Agent interactions.

**Voice Model Class**

A grouping of related voice models by capability or quality tier.

**Voice Model Provider**

The vendor that hosts a given voice model.

**Voice Persona**

The spoken identity—tone, style, and voice selection—used by a voice-enabled Agent.

## W

**Warning**

A cautionary notice shown in the UI or logs when a condition needs operator attention.

**Web Page**

A Source Type for content ingested from a URL on the web.

**Web Scraping**

Automated extraction of web page content during source import or indexing.

**Webhook**

An HTTP callback Apologist sends to your endpoint when subscribed events occur.

**Webhook Request Signing**

Cryptographic verification—typically HMAC—that a webhook payload came from Apologist.

**Webhook Secret**

A shared secret used to sign or validate webhook requests.

**Website**

A Collection Type representing pages gathered from a site for ingestion and retrieval.

**WhatsApp**

A Channel integration that lets users converse with an Agent on WhatsApp.

## Y

**YouTube Channel / Playlist**

A Collection Type for video content sourced from a YouTube channel or playlist.

**YouVersion**

A Bible Translation Provider that supplies YouVersion scripture content for Agent use.