Helpin
Dark mode
Translation quotas & settings

Translation quotas & settings

Live Translate settings live in support settings when the feature is enabled. They control defaults, cost protection, and which languages participate.

Beta. Live Translate is in beta. This page describes the settings available once it is enabled for your workspace.

Workspace settings

Setting

Type

Default

Purpose

Enable live translate

bool

off

Master switch for the workspace.

Auto-suggest for language mismatch

bool

off

Suggest enabling translation when a mismatch is detected.

Default agent language

language tag

workspace locale or en

The agent-side language used for inbound translations.

Monthly character limit

characters/month

500,000

Workspace-wide translated-character budget per calendar month.

Request rate limit

requests/window

120 per 60s

Rolling-window cap on translation requests.

Excluded languages

list

empty

Languages never translated (for example, languages your team speaks natively).

Store originals

bool

on

Whether original inbound/outbound text is retained with translations for audit and reveal.

Detection minimum confidence

0.0–1.0

0.75

Below this, translation is not auto-suggested.

Users with translation-manage permission can edit the quota and rate limit; all users can use translation per conversation.

Quotas and rate limits

Translation usage is tracked and enforced:

  • Monthly characters — every translated character (inbound and outbound, human and AI answers) counts. Helpin checks the quota before translating; if a request would exceed the limit, it is not translated and no text is sent for translation.

  • Requests per window — a rolling cap on request volume to protect stability.

When limits are hit:

  • The conversation header shows "Translation quota reached".

  • Agents can always send without translation — normal replies are never blocked by translation state.

  • Usage resets at the start of the next calendar month.

Caching

Identical translations are served from cache:

  • Repeated display of the same message (another agent opening the conversation, page reloads) reuses the existing translation instead of translating again.

  • This makes repeat views fast and cheap; only genuinely new text consumes quota.

Usage visibility

  • Each translation and language detection is recorded with its character count and outcome (success, translation error, quota reached, rate limited).

  • A usage summary in settings shows current-month consumption against the limit.

  • Alerts fire as the workspace approaches quota (90% and 100%).

Privacy and security

  • Message text is sent to the translation provider only when translation is active for that message; obvious secrets (API keys, bearer tokens, passwords, card-like numbers) are redacted before provider calls where safe.

  • Raw message text is never logged; structured logs carry only metadata (languages, character counts, latency, status).

  • Originals are stored only if Store originals is on, and retention follows conversation retention.

Was this article helpful?