Help

Glossary

Every word the portal and these docs use with a meaning of its own, in alphabetical order, one or two sentences each, with the page that explains it in full.

The portal and these docs use a small set of words with a fixed meaning. This page defines each one in a sentence or two and links the page that explains it in full. Look a word up here when a page uses it before it explains it, or when two pages seem to use it differently.

A

Agent. One saved configuration that answers people on your behalf. It holds a persona, a model, one or more voices, the knowledge it may quote from, the tools it may call, and the channels it can be reached on. An idle agent costs nothing; a conversation is charged for its connected time and its AI usage. What is an agent.

Allowance. The plan credits a workspace receives each month, shown on the plan card as credits every month, and the knowledge storage the plan grants. Unused credits roll over up to three times the allowance; the storage is a ceiling the workspace stays under. Credits, top-ups and rollover.

API key. The secret your own backend sends as a bearer token on every Developer API request. It starts with ek_, is shown once when created, and belongs to one workspace. Get an API key.

Archived. The status an agent holds after Pause Agent. An archived agent cannot be edited or tested from the portal and does not count toward the plan's agent limit; Restore makes it active again. Agent status.

Assets. The page that records every Text to Speech generation and every Speech to Text transcript made from the portal in the workspace, newest first, with play, download and recreate on each item. Assets and history.

Auth connection. A saved credential with an authentication method attached, basic, bearer or OAuth2, that a tool uses to produce its own Authorization header. Save it once and attach it to every tool that calls that API. Auth connections.

Auto top up. A rule that buys a batch of extra credits for the workspace whenever the balance falls below a threshold you set, charged to the card on the subscription, within a monthly limit you set. Credits, top-ups and rollover.

B

Basic limits. The limits a workspace has with no paid plan, after a cancellation or an unpaid renewal: no plan credits, the Free Trial plan's feature limits, and extra credits still usable. The Subscription page reads No paid plan, basic limits. Credits, top-ups and rollover.

Binding. The link between one channel and the one agent that answers on it. A connected account is bound to exactly one agent; Assign to another agent moves the binding, and the next message opens a conversation with the new agent. Channels overview.

C

Channel. A place your customers already are, connected to one of your agents: a WhatsApp number, a Slack workspace, a phone number, a helpdesk account. Every message or call that arrives on it becomes a conversation with that agent. Channels overview.

Chunk. One passage a knowledge source was split into when it was indexed. The Chunks column shows how many a ready source has; a very low count means very little text could be read from it. How knowledge works.

Conversation. One call or one chat an agent handled, kept with its transcript and the numbers measured while it ran. Every conversation belongs to exactly one agent and is listed on that agent's Conversations tab. Find and read a conversation.

Conversation tag. A label the workspace curates and puts on conversations. Tags belong to the workspace, not to an agent, so one tag can sit on conversations from every agent you run. Tag a conversation.

Credit. The unit usage is paid in, from one balance. A generation, a transcription or a cleanup costs a fixed number of credits, taken before the work runs and returned if it fails; AI usage is the provider's price plus 10%, turned into credits. Core concepts.

Credit limit. An optional ceiling on one API key: the credits it may spend per billing cycle. A Text to Speech call that would cross it is refused with 429 Monthly spending limit exhausted. Get an API key.

Credits History. The list on the Payment History page that shows every credit consumed or refunded: what each speech job cost, whether it was refunded, and the balance after it. Track your credit usage.

D

Database connection. A read-only connection to your own database, registered once so an agent can answer callers from live data. It is not a row in the tool library; the agent reaches it through the questions you approve. Connect a database.

Database question. One read-only query, drafted from your schema and checked against your database, that the agent may run with parameters it fills in. Nothing reaches a conversation until you approve it. Review and approve database questions.

Developer API. The HTTPS API your own systems call for Text to Speech and Speech to Text, authenticated with an API key and billed in credits against the workspace. Developer API overview.

Dialect. The accent a voice speaks in. Every voice speaks exactly one, so you choose a voice and its dialect comes with it; Fusha is Modern Standard Arabic and the others are the speech of one country. Dialects.

Diacritics (tashkeel). The marks that fix how an Arabic word is pronounced. The Tashkeel switch on Text to Speech marks a passage before an Arabic voice reads it, keeping any marks you typed yourself. Add diacritics.

Diacritization. The Tashkeel control, ON or OFF, on an agent's voice that adds diacritics to the text before it is spoken. It appears on every voice except English ones. Voices and language.

E

Extra credits. Credits bought with a top-up, by you or by auto top up. Each batch never expires, survives a plan change or a cancellation, works with no paid plan, and is spent only when the plan credits cannot cover a charge. Credits, top-ups and rollover.

F

Folder. A group of sources inside the knowledge library. Selecting a folder for an agent is a live reference, so anything you drop into that folder later reaches the agent with no further change. Organize knowledge in folders.

Fusha. Modern Standard Arabic, understood across the region. It is the dialect the API's tashkeel option applies to. Dialects.

G

Generation. One piece of speech made from text on the Text to Speech page or through the API. Every generation made in the portal is kept in the History panel and on Assets until you delete it, and replaying it costs nothing. Text to Speech history.

Greeting message. The first line the agent says when a conversation starts, written on the Persona tab. It may be left empty. Write the persona.

H

Handoff. The moment the agent stops and a person takes over. On Zendesk, Zoho Desk, Intercom and Genesys Cloud chat the agent marks the ticket or conversation for your team and goes quiet; on every other channel the persona has to carry the customer to a person in words. Human handoff.

I

Integration. A provider you connect once so that its ready-made actions become tools your agents can call, such as the calendar tools of a connected Google account or Send SMS on a connected Twilio number. Integration tools carry their own credentials, so there is nothing to configure per call. Tools overview.

J

Job. One asynchronous Text to Speech or Speech to Text request on the API. A job moves through four statuses, pending, processing, completed and failed, and you read its result from the job endpoint or through a webhook. Synchronous and asynchronous calls.

K

Knowledge source. Material an agent can quote from: a file you upload, text you paste, or a web page or site you point the library at. Sources belong to the workspace, each agent picks the ones it may use, and only a source at Ready is used. How knowledge works.

M

Model. The one model that thinks for an agent, chosen on the Voice tab under Model. On models that support it you also set how hard it reasons before it answers. Choose a model and reasoning effort.

N

Notification. A message the portal sends you in the bell in the header: a knowledge source finishing or failing, a payment succeeding or failing, a voice session ending for lack of credits. Read your notifications.

O

OpenAPI tool. One imported Swagger or OpenAPI document. The whole API becomes a single row in the tool library, and the agent finds the right operation inside it during a conversation. OpenAPI tools.

Operation. One record inside an OpenAPI tool, one per operation in the imported document, with its own auth mode, fixed values and enable switch. OpenAPI tool operations.

P

Persona. The tab that holds who the agent is in your workspace, how it speaks, and the first line it says. It changes an agent's behaviour more than any other tab. Write the persona.

Personality. The prompt. It is the one box the model reads word for word, and every wording rule an agent follows comes from it: its identity, its scope, its tone, what it refuses and when it hands over. Write the persona.

Plan. The subscription a workspace runs on: a trial with no time limit to begin with, then the paid plan you choose on monthly or yearly billing. The plan card lists its credits, agents, storage, API keys and services. Choose or change your plan.

Plan credits. The credits a plan grants every month. They are spent first, and what is left rolls over when the same plan renews, up to three times the allowance. Credits, top-ups and rollover.

Playground. A real conversation with your agent started from the portal with Test, by Chat or by Call, instead of from a channel. It runs the same agent a customer would reach and is charged the same way. Test your agent.

Preview. The short clip most voices in the library carry, played from the orb on the voice row so you can hear the voice before you use it. Browse the voice library.

R

Reasoning effort. How hard the model reasons before it answers, set on the Voice tab for models that support it: Default, Minimal, Low, Medium or High. Higher effort can improve quality and adds latency. Choose a model and reasoning effort.

Recovery code. The one code you save when you turn on two-factor authentication, for signing in when the authenticator app is not to hand. Two-factor authentication.

Relevance threshold. The retrieval setting that decides how closely a passage must match the turn before the agent may use it. The agent keeps the passages that clear it, up to Top documents of them. How knowledge works.

Request log. The list under Developers on the Request Log tab of every POST request that passed authentication on the Developer API, with its status and the key that made it. Read the request log.

Retrieval. The step, on every turn of an agent with Enable retrieval on, that searches the sources the agent may use and puts the matching passages in front of the model. With retrieval off, every selected source is ignored. How knowledge works.

Retrieval bench. The Test Tool check on an imported OpenAPI tool that answers one question: would this agent reach this operation if a caller said that? It is a retrieval test, not an execution test. The retrieval bench.

Rollover. Unused plan credits carried into the next month when the same plan renews, up to three times the monthly allowance. Extra credits do not count toward the limit. Credits, top-ups and rollover.

S

Secret. One named value that belongs to the workspace, sealed on the way in, never shown again, and referenced by name from a tool's header. Use it for an API key; for a username and password, a bearer token or OAuth2, use an auth connection. Workspace secrets.

Session. One run of the playground, from Chat or Call until you close the overlay. Closing always ends the session, and everything said in it lands on the agent's Conversations tab like any other conversation. Test your agent.

Signing secret. The value, starting with whsec_, that every new webhook receives once so your server can verify each delivery. The portal and the API keep only its first 12 characters afterwards. Configure a webhook.

Summary. The short account of a conversation written once, at the end, when the agent wraps the conversation up. A conversation can have a complete transcript and no summary. Transcripts and summaries.

Synchronous and asynchronous. The two call styles of the Developer API. A synchronous call holds the connection open and returns the result; an asynchronous call returns a job id at once and does the work in the background. Synchronous and asynchronous calls.

System tool. One of the three built-in capabilities every agent has, switched off until you switch it on: web search, dialect detection and end conversation. They live on the Tools tab under Built-in tools and are never added or deleted. System tools.

T

Top documents. The retrieval setting that caps how many matching passages the agent keeps on one turn, once they clear Relevance threshold. How knowledge works.

Top-up. A purchase of extra credits with Add More Credits on the Subscription page, on a paid plan or on a trial. Each top-up is its own batch and never expires. Credits, top-ups and rollover.

Transcript. What a conversation records as it runs: one message for each side of every turn, with the numbers measured on that turn. On the Speech to Text page, the text produced from an audio file. Transcripts and summaries.

Trial. The subscription every new workspace starts on, with 60,000 credits, no time limit and up to $1 of AI model cost. Subscribing to a paid plan ends it, and the unused trial credits end then. Choose or change your plan.

Turn. One exchange in a conversation: the customer says something, the agent answers. Each turn charges its AI usage, and on a call the connected time since the last turn. What a turn costs.

Two-factor authentication. A second check on every sign-in: after your password, or after Google or Microsoft returns you to the portal, you enter the 6-digit code your authenticator app shows for EchoNova. Two-factor authentication.

V

Voice. What speaks. The library under Voices holds every voice you can use, the same in every workspace; each has a name, a dialect and a style line, and reads either the text you type on Text to Speech or what an agent composes in a conversation. Browse the voice library.

W

Webhook. An HTTPS endpoint you own that EchoNova calls when an asynchronous job finishes or an API key crosses its credit limit. You create it in the portal, choose its events and keep its signing secret. Configure a webhook.

Webhook tool. One HTTP endpoint of yours, described once with a name, a description the agent reads, a URL and parameters, that the agent can call mid-conversation with your credentials already on the request. Webhook tools.

Working hours. An agent's time zone and answering window, set on the Settings tab. They record when you intend the agent to be available; to stop an agent answering, disconnect its channels or bind them to another agent. Working hours.

Workspace. The container everything you build lives in: agents, knowledge, tools, conversations, speech generations, transcripts, API keys and the subscription. One is created for you at sign-up with you as Owner. Workspaces.

Workspace tools. The section of an agent's Tools tab that lists the webhook, OpenAPI and integration tools in the workspace library, which any agent in the workspace can select. Tools overview.

FAQ

Is a voice, a knowledge source or a tool mine or the workspace's?

The voice library is the same in every workspace. Knowledge sources, tools, secrets, auth connections, conversation tags and agents belong to the workspace; your account holds only your name, sign-in method, two-factor setting, theme, language and notifications. Workspaces.

What is the difference between a secret and an auth connection?

A secret is one opaque value you name a header for. An auth connection knows how to authenticate on its own and always produces the Authorization header itself. Auth connections.

What is the difference between plan credits and extra credits?

Plan credits arrive each month with the plan, are spent first, and roll over up to three times the allowance while the subscription is active. Extra credits come from top-ups, never expire, and are spent only when the plan credits cannot cover a charge. Credits, top-ups and rollover.

Is the playground a channel?

No. Web chat and web calls from Test need no channel; they run the same agent a customer would reach and are charged the same way. Test your agent.