collapsible details
tool arguments and results collapse by default. click to expand. keeps the chat clean while preserving full context.
tools
every tool state handled — pending, in-progress, completed, failed, and awaiting human approval. inline, collapsible, with rich results.
pnpm dlx shadcn@latest add @inferencesh/toolspending, in-progress, completed, failed, awaiting approval, and widget results. one component handles the full lifecycle. collapsible details keep the chat clean while preserving full context.
{
"query": "inference shell pricing"
}{"urls":["https://inference.sh/pricing"],"snippets":["pay per second, no minimums"]}{
"path": "/data/old-backup.tar.gz"
}wrap in AgentChatProvider and pass the invocation DTO. the component handles status indicators, collapsible arguments and results, approval buttons, and widget rendering automatically.
import { ToolInvocation } from '@inference/common/agent' import { AgentChatProvider } from '@inferencesh/sdk/agent' // wrap in AgentChatProvider, then render each invocation <AgentChatProvider client={client} agentConfig={config}> <ToolInvocation invocation={toolCall} defaultOpen={true} /> </AgentChatProvider>
features
tool arguments and results collapse by default. click to expand. keeps the chat clean while preserving full context.
tools that require human approval render inline approve/reject buttons. no modal interruptions, no context switching.
tool results can render as widgets — charts, tables, cards — instead of raw JSON. structured output, rich UI.
when a tool invokes an inference.sh app, the task output renders inline with streaming progress and status.
start with the hosted platform. deploy your own when you're ready.
we use cookies
we use cookies to ensure you get the best experience on our website. for more information on how we use cookies, please see our cookie policy.
by clicking "accept", you agree to our use of cookies.
learn more.