Glossary
Definitions of platform-specific terms. Alphabetical. If a term has its own dedicated article, the entry links there.
A
Active model The currently selected model in your chat. Visible in the top bar; click to change. Different from your project's preferred model (which is the default for new conversations in that project). See Picking a Model.
Active project The project you're currently inside. The chat header shows its name. While active, all new conversations go into it and the project's system prompt prepends every message. See Project-Scoped Chat.
Agent Mode Multi-step autonomous mode where the model uses tools (web search, quote lookup, etc.) to work toward a goal. Different from regular chat which is single-turn. See What is Agent Mode?.
Agent run A single execution of an agent goal. Has a unique ID, a sequence of steps, a final answer, and a cost. See Quick Start.
Apsis The sci-fi chassis. Glassmorphism + parallax starfield + indigo/cyan accent. Calls projects "Missions" and conversations "Transmissions." See Welcome to Apsis.
Artifact A user-saved or auto-generated piece of content (text, markdown, code, JSON, list) stored in your local IndexedDB vault. Different from chat messages or memories. See What Are Artifacts?.
Artifact vault Your personal collection of saved artifacts. Local to your browser. Searchable, taggable, versioned. See Vault Overview.
Atelier The artist-studio chassis. Warm tones, gallery layout, vermillion accent. Calls projects "Collections" and conversations "Pieces." See Welcome to Atelier.
Atlas The exploration / education chassis. Cobalt + highlighter yellow, journal styling. Calls projects "Courses" and conversations "Lessons." See Welcome to Atlas.
Auth Short for authentication. Nyquest uses Bearer tokens (JWT for sessions, PATs for scripts). See Authentication.
Auto-router The default routing mode. Scores each message for complexity and domain, then picks a model by tier, live price, and demonstrated per-domain quality. See Auto-Router: Tiers, Domains, and Providers.
B
Bearer token
The auth header format Nyquest accepts: Authorization: Bearer <token>. Token can be a JWT or a PAT. See Authentication.
BYOK "Bring Your Own Key." Add your provider's API key (OpenAI, Anthropic, etc.) and your requests go through your account, not Nyquest's. Free on the platform side. See BYOK Setup.
C
Case What Ridgeline chassis calls a project. See Per-Chassis Terminology.
Chassis One of 10 complete UI design languages. Each chassis has its own visual identity, terminology, and chassis-specific features. Picking a chassis = picking your work mode. See Picking a Chassis.
Chassis Welcome Per-chassis intro article (one for each of the 10) explaining that chassis's panel layout, terminology, and what it's good for. See Welcome to Ridgeline etc.
Chat completions
The OpenAI-compatible endpoint at POST /v1/chat/completions. Standard request/response shape. See OpenAI Compatibility.
Circuit breaker Automatic protection around each upstream provider: a few consecutive failures take the provider out of rotation briefly while traffic fails over, to reduce disruption. Failures can still reach callers; provider-specific routing features depend on deployment configuration. See Auto-Router: Tiers, Domains, and Providers.
Collection What Atelier chassis calls a project. See Per-Chassis Terminology.
Compression The pipeline that shrinks prompts before they're sent upstream, cutting token cost without changing meaning. Runs on managed and BYOK chat, and once-before-fan-out on Splicer runs. See Multi-Model Splicer.
Concurrent run limit The number of simultaneous runs. Nyquest does not publish a Free-3/Pro-10 concurrency allowance; Agent starts share chat request limits and every planner call is subject to wallet and run-budget checks. See Rate Limits and Quotas.
Consensus Guard The platform's continuous quality loop: a sample of routed traffic is quietly fanned to alternate models and judged on multiple axes; the results feed the auto-router's model rankings. The Splicer is the same machinery under your manual control. See Multi-Model Splicer.
Conversation A series of messages between you and a model, with shared context. Auto-saved. May be inside a project or top-level. See Conversations and History.
Course What Atlas chassis calls a project. See Per-Chassis Terminology.
Courtroom The legal chassis. Oxford blue + parchment + serif. Calls projects "Matters" and conversations "Briefs." See Welcome to Courtroom.
D
Deep link
A URL that points directly to a specific help article (e.g. /help/agent-mode/quick-start). Slash command /help <category>/<slug> opens any of these.
Developer API
The HTTP API at https://api.nyquest.ai. See Developer API Overview.
Domain (routing) The subject-matter class the auto-router detects for each message — Code, Creative, Reasoning, Vision, or General — used to pick a model with demonstrated strength in that area. See Auto-Router: Tiers, Domains, and Providers.
E
Embedding A vector representation of text used for semantic search. Memory recall and artifact search use embeddings. See Recall and Surfacing.
F
Fact A discrete claim extracted from your conversations and stored for recall. "User prefers Rust" is a fact. See What Gets Remembered.
Folio What Paper chassis calls a project. See Per-Chassis Terminology.
Free mode How anonymous visitors and $0-balance accounts get served: a self-healing chain of genuinely free catalog models, at no cost. Different from the Free tier (an account level). See Free Mode.
Free tier Default account tier. 20/min chat rate limit, 10MB per artifact, no live support access. Funding the wallet (any amount) flips you to Pro. See Plans and Pricing.
Fuzzy search Substring text search across titles + bodies + tags. Default search mode in the artifact vault. Different from semantic search. See Search and Tags.
G
Garden The wellness chassis. Sage + sand, breathing animation, calm pace. Calls projects "Gardens" and conversations "Seedlings." Sign-out is "Farewell." See Welcome to Garden.
H
HelpButton
The ? icon in chassis topbars and panel headers that opens the help section. Reusable component. Accepts a topic prop for deep-linking.
HelpHome
The landing page at /help. Shows category cards + chassis-aware welcome CTA + search bar.
I
Idle session A live support session with no activity for 30+ minutes. Marked "idle" automatically. After 24h of inactivity, auto-closed. See Contacting Support.
IndexedDB The browser-side database where artifacts are stored. Local to your browser; never on Nyquest's servers. See What Are Artifacts?.
is_pro
Database flag indicating Pro tier. Flips to true after first wallet funding, never reverts.
J
JWT JSON Web Token. The session token issued at sign-in. Configured lifetime: 8 hours. Different from a PAT (long-lived). See Authentication.
L
Lesson What Atlas chassis calls a conversation. See Per-Chassis Terminology.
Live support
Discord-bridged operator chat at /help/live-support. Pro users only. See Contacting Support.
localStorage
Browser key/value store. Nyquest uses it for chassis pick (nq_chassis), active project (nq_active_project), JWT (nq_token), etc. Per-browser, not synced.
M
Mandate What Quant chassis calls a project. See Per-Chassis Terminology.
Matrix The terminal chassis. True black + phosphor green + falling code rain. Calls projects "Ops" and conversations "Sessions." See Welcome to Matrix.
Matter What Courtroom chassis calls a project. See Per-Chassis Terminology.
MCP (Model Context Protocol) A protocol for AI tools to discover and call services. Some external MCP servers can integrate with Nyquest via the Chrome extension; the Nyquest backend itself is not an MCP server. See Webhooks and Integrations.
Memory recall The platform's automatic injection of relevant past context (facts, summaries, embeddings) into your current message. See Recall and Surfacing.
Mission What Apsis chassis calls a project. See Per-Chassis Terminology.
N
nq-v1-
Prefix on every personal API key. Shorthand for "Nyquest personal access token." See Personal API Keys.
O
Op What Matrix chassis calls a project. See Per-Chassis Terminology.
OpenAI-compatible
The shape of POST /v1/chat/completions matches OpenAI's, so any OpenAI client works against Nyquest with just a base URL change. See OpenAI Compatibility.
P
Paper The document-first chassis. Warm white, serif throughout. Calls projects "Folios" and conversations "Notes." See Welcome to Paper.
PAT
Personal Access Token. Long-lived API key for scripts. Created at /user/api-keys, never expires. See Personal API Keys.
Pin (artifact) Marking an artifact for prominent display at the top of the vault list. Visual only — doesn't change recall priority.
Planner The model that drives an agent run's step decisions. You can pick which model serves as planner per-run. See Quick Start.
Preferred model A project's pinned model. New conversations in the project default to this model unless overridden. See Creating a Project.
Project A folder for related conversations sharing a system prompt + preferred model. Per-chassis terminology. See What is a Project?.
Pro tier Account tier after first successful wallet funding (minimum $5 through Checkout). 120/min chat rate limit, live support, 100MB per artifact. Pro also enables the Multi-Model Splicer. See Plans and Pricing.
Provider layer An optional routing layer that picks which upstream provider serves a model. It is currently disabled on the hosted deployment; do not assume provider-by-provider automatic failover. Compare providers at Settings → Benchmarks; steer with the Preferred Provider setting. See Auto-Router: Tiers, Domains, and Providers.
Q
Quant The trading-desk chassis. Deep navy + gold, scrolling ticker, tabular figures. Calls projects "Mandates" and conversations "Memos." See Welcome to Quant.
Quote lookup Agent Mode tool that fetches live equity / crypto / FX prices from Stooq. Free, cached 30s. See The 6 Tools.
R
Rate limit Cap on requests per minute. 20/min free, 120/min Pro for chat. 5/min per IP for auth. See Rate Limits and Quotas.
Recall chunk A semantically-indexed chunk in the memory layer. Built from facts, summaries, and selected message bodies. The unit recall searches over. See Recall and Surfacing.
Repo What Studio chassis calls a project. See Per-Chassis Terminology.
Ridgeline The default chassis. Light theme, hairline borders, medical teal. Calls projects "Cases" and conversations "Charts." See Welcome to Ridgeline.
S
Semantic search Embedding-based similarity search. Slower than fuzzy but finds conceptually related content even without exact words. See Search and Tags.
Session (live support) A user-operator chat thread, Discord-bridged. Auto-closes after 24h idle. See Contacting Support.
Session (Matrix/Studio chassis) What those chassis call a conversation. See Per-Chassis Terminology.
Slash command
A command typed in chat input starting with / that's intercepted before sending. /help, /help <category>, /help <category>/<slug> are the current set. See Slash Commands.
Splicer Sends one prompt to several models in parallel and judges the answers for consensus: agreement %, a synthesized answer, and divergences. Pro and up. See Multi-Model Splicer.
SSE (Server-Sent Events) HTTP-based streaming protocol. Nyquest uses SSE for chat completion streaming, agent run events, and live support events. Different from WebSockets.
Streaming Token-by-token output as the model generates. Default for chat. Visible as words appearing live. See Streaming.
Studio The recording-studio chassis. Charcoal + magenta + lime, mixer-style layout. Calls projects "Repos" and conversations "Sessions." See Welcome to Studio.
Summary (memory)
A short summary of a completed conversation, stored in memory_summaries and embedded for recall. See How Memory Works.
Superseded fact A fact that's been replaced by a newer one ("user moved to Berlin" supersedes "user lives in NYC"). Excluded from recall but kept on record. See What Gets Remembered.
System prompt Text prepended to every message in a project. The project's most powerful customization lever. See System Prompts.
T
T0 (local tier) The $0 routing tier served by a model on Nyquest's own hardware. Handles simple auto-routed tasks, streaming included, with clean fallback to hosted tiers. See Auto-Router: Tiers, Domains, and Providers.
Tag A short label on an artifact for filtering. Up to many per artifact, comma-separated entry. See Search and Tags.
Tier (T0–T4) The price bands the auto-router maps tasks onto, from T0 (local, $0) through T4 (frontier models). Tier membership comes from live catalog prices, not model names. See Auto-Router: Tiers, Domains, and Providers.
Token Units the model counts (roughly: ¾ of a word in English). Pricing is per-million-tokens. See Picking a Model.
Tool (agent) A capability the agent can invoke during a run. Currently 6: web_search, web_fetch, quote_lookup, memory_recall, artifact_read, project_metadata. See The 6 Tools.
Top-K (recall) The number of recall chunks pulled in per message. Default 3-5. See Recall and Surfacing.
Transmission What Apsis chassis calls a conversation. See Per-Chassis Terminology.
V
Version (artifact) Each edit to an artifact creates a new version automatically. Full version history is queryable; old versions are restorable. See Vault Overview.
W
Wallet Your account's prepaid balance. Funded via Stripe Checkout in increments. Debited per platform-hosted request. See Wallet and Funds.
Watchlist (Quant ticker) The list of symbols shown in the Quant chassis ticker. Configurable via the gear icon in the ticker bar. Per-browser. See Visual Effects.
Welcome tour
The 4-step modal shown on first sign-in. Briefly orients the user; can be skipped; sets nq_welcome_tour_seen so it doesn't reappear.
Webhook HTTP callback. Nyquest doesn't currently push outbound webhooks; the only inbound is the Stripe billing webhook. See Webhooks and Integrations.
Where to next
- FAQ — common questions
- What's New — recent shipped features
- Help home — browse all categories