Read these FAQs to quickly understand Nexa basics and recommended prompt guidelines.
-
Does token consumption vary based on the time of day requests are made?
The platform applies uniform token usage across all hours of the day. No off-peak pricing tier or token discount exists; consumption rates remain consistent regardless of when requests are executed.
-
What happens when a customer reaches their monthly token quota—does the platform permit overages?
Each customer receives a monthly allocation of 50 million tokens. Before any request executes, the platform performs a pre-request quota check to validate the available token balance. When a customer's monthly limit is reached, the platform blocks the incoming request and returns a rejection response. No overages are permitted; usage does not exceed the defined cap.
-
Are tokens consumed during a failed NEXA query refunded?
When NEXA is unable to return results, the tokens consumed during that request are not refunded. Token usage is recorded at the point of request execution, irrespective of whether the query produces a successful result.
-
When a user reopens a previous conversation, does Nexa rerun the prompts or retrieve them from storage?
When a user reopens a previous conversation, Nexa retrieves conversation events from storage. These stored events are not rerun and do not consume additional tokens. However, any new prompt that loads context within the resumed conversation does consume tokens at the standard rate.
-
When a conversation is shared, are the prompts rerun for the recipient, or is a cached version presented?
Shared conversations present a cached version of the conversation state. No prompts are rerun during the sharing process. Recipients access the conversation as it was captured, without triggering additional model execution.
-
How long are shared conversations available?
Shared conversations remain available for 30 days from the time of sharing. After this period, access to the shared conversation expires.
-
Does Nexa retain business-specific terminology or context introduced in one conversation for use in future conversations?
Nexa does not currently support memory across conversations. Each conversation context is constructed from the configured tools, the knowledge base, and — in V2 — Skills, along with any context the user explicitly provides within that conversation. Terminology or instructions introduced in one conversation, such as business-specific terms, do not persist into subsequent conversations.
A permanent, per-client context layer — ingested at the client level — is under consideration for a future release. This capability would allow business-specific knowledge to persist across conversations, producing more contextually relevant outputs for each client.