Skip to main content
Gremorie

Assistant

The flagship Gremorie chat surface — Conversation, Message, Reasoning, an inline Artifact, Sources and a B2B PromptInput composer, fully composed.

Overview

The Assistant is the flagship Gremorie block: a production-ready chat surface with a thread switcher and actions header, a streaming conversation (reasoning, an inline generated artifact, sources, response branches with actions), and a floating B2B PromptInput composer — context picker, a token-usage meter, mode and model selectors, web search, attachments, voice and a status-aware submit.

Wire onSubmit to the AI SDK or any streaming endpoint and you get a full assistant experience in a single import.

Preview

q3-revenue.jpg
Summarize our Q3 revenue and draft a one-pager I can share with the board.

Pulling Q3 revenue from the finance warehouse, comparing against Q2 and the forecast, then drafting a board-ready one-pager. Keeping it to headline, three KPIs, and one risk.

Here is the board one-pager. Q3 revenue landed at $4.2M, up 18% QoQ and 6% ahead of forecast, driven by net expansion in the enterprise tier.

Headline KPIs

  • ARR: $16.8M (+18% QoQ)
  • Net revenue retention: 114%
  • Gross margin: 78%

Watch item: enterprise pipeline coverage for Q4 sits at 2.9x, below the 3.2x target. The full one-pager is attached below.

Q3 one-pager

Document · generated just now

Q3 closed at $4.2M, up 18% QoQ and 6% ahead of forecast.

MetricQ3QoQ
Revenue$4.2M+18%
ARR$16.8M+18%
NRR114%+3pp

Gremorie can make mistakes. Check important info.

Anatomy

PartRole
AssistantHeaderThread/history switcher (left); Sources, Share and an overflow menu (right)
AssistantContentThe conversation — messages with Reasoning, an inline generated Artifact, Sources, and response MessageBranch + MessageActions
AssistantComposerThe floating PromptInput (B2B): PromptInputContext, the Context usage meter, mode + model selectors, web search, attach, voice, and a status-aware submit
AssistantDisclaimerThe "can make mistakes" footnote

Installation

npx gremorie@latest add block-assistant
pnpm dlx gremorie@latest add block-assistant
yarn dlx gremorie@latest add block-assistant
bunx --bun gremorie@latest add block-assistant

Composition

The block composes (condensed — the full, runnable source installs via the CLI above):

<TooltipProvider>
  {/* AssistantHeader — thread switcher + actions */}
  <Conversation>
    <ConversationContent>
      <Message from="user">…</Message>
      <Message from="assistant">
        <MessageBranch>
          <Reasoning />
          <MessageResponse />
          <Artifact /> {/* inline generated document */}
          <Sources />
          <MessageToolbar>{/* actions + branch selector */}</MessageToolbar>
        </MessageBranch>
      </Message>
    </ConversationContent>
  </Conversation>

  {/* AssistantComposer — the floating B2B PromptInput */}
  <PromptInput>
    <PromptInputHeader>
      <PromptInputContext />
      <Context>
        <ContextTrigger />
        <ContextContent>{/* usage breakdown + cost */}</ContextContent>
      </Context>
    </PromptInputHeader>
    <PromptInputBody>
      <PromptInputTextarea />
    </PromptInputBody>
    <PromptInputFooter>
      <PromptInputTools>{/* mode + model selects */}</PromptInputTools>
      <PromptInputTools className="gap-2">
        <PromptInputTools>{/* search · attach · voice */}</PromptInputTools>
        <PromptInputSubmit />
      </PromptInputTools>
    </PromptInputFooter>
  </PromptInput>
</TooltipProvider>

Customization

  • Wire onSubmit to the AI SDK's sendMessage or your streaming endpoint
  • Stream Reasoning while isStreaming is true for auto-collapse on completion
  • The composer is the B2B PromptInput — adjust the PromptInputContext items, the mode/model presets, and the Context usage meter
  • Swap MessageContent for MessageResponse (Streamdown) for server-streamed markdown

On this page

OverviewPreviewAnatomyInstallationCompositionCustomizationRelated