Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
import type { SessionTokensLite } from '../../hooks/useChatStore'
import { formatTokenBreakdown } from '../../lib/format'
import { Pill } from '../Common'
import { TokenUsageStats, TokenUsageTooltip } from './TokenUsage'
export interface SessionTokensIndicatorProps {
tokens: SessionTokensLite
}
/**
* Discreet header pill showing the active conversation's lifetime token usage
* as separate input (↑) / output (↓) compact stats, plus a cache glyph when
* the model reports cache figures. The exact breakdown lives in the custom
* hover/focus tooltip (see TokenUsage). Updates on re-poll, like `turnCount`.
*/
export function SessionTokensIndicator({ tokens }: SessionTokensIndicatorProps) {
if (tokens.input + tokens.output <= 0) return null
return (
<div className="agent-chat-tokens-row">
<Pill
tone="neutral"
size="sm"
className="token-usage token-usage--pill"
tabIndex={0}
aria-label={`Conversation token usage — ${formatTokenBreakdown(tokens)}`}
>
<TokenUsageStats tokens={tokens} />
<TokenUsageTooltip tokens={tokens} />
</Pill>
</div>
)
}