Interface ChatWidgetProps

Accent color override; wins over the theme's accent in both light and dark.

Absolute URL of the Worker chat endpoint (e.g. https://api.example.com).

Let visitors attach images and PDFs (click, drag-and-drop, or paste). true enables the defaults (5 MB per file, 5 files per message, JPEG / PNG / GIF / WebP / PDF); pass an AttachmentsOptions to tune them. The worker must allow attachments too (it does by default).

Let visitors copy the conversation as Markdown, or download it as Markdown or JSON, from a menu in the chat header. Everything happens in the browser and nothing is sent to the Worker. Only the literal true enables it. See the Conversation export guide.

BCP-47 locale used to select built-in translations.

Persist the conversation to storage so it survives reloads.

Placeholder text for the message input.

Middleware run around each message: onBeforeSend, onAfterReceive, and onError. Hooks run in array order and may modify, replace, or short-circuit messages. See ClaudiusPlugin.

Corner of the viewport the widget docks to.

Abort a chat request after this many milliseconds.

Prefix for the storage key used when ChatWidgetProps.persistMessages is enabled.

Stream assistant replies token-by-token from the Worker's SSE endpoint, with automatic fallback to the non-streaming API on older browsers or Workers without /api/chat/stream. Set to false to always use the blocking request/response flow.

Header subtitle shown beneath the title.

Color-scheme mode ("light" | "dark" | "auto"), a built-in theme name ("default" | "minimal" | "playful" | "corporate"), an inline ClaudiusTheme object, or a URL to a theme JSON file.

Header title. Falls back to the active locale's default title.

Partial overrides merged over the resolved locale translations.

Proactive open/greeting rules evaluated against the current page.

Voice features built on the browser's Web Speech API: a mic button for dictating messages and a read-aloud control on assistant replies. true enables both (click-to-toggle mic, no auto-submit); pass a VoiceOptions to choose hold-to-talk, auto-submit, a language, or to switch either half off. Each half renders only where the browser supports it. The widget never handles audio itself, but the browser's speech service may process it remotely; see the Voice guide.

First assistant message shown when the chat opens.