The agent that grows with you
Python aiai-agentai-agentsanthropicchatgptclaudeclaude-codecodexhermeshermes-agentllmnous-researchopenai
/**
* What a bot's chat shows before it has said anything.
*
* Core's splash is Hermes' own wordmark and belongs to a fresh draft; a bot
* chat is neither. It gets the same lettering with the bot's name in it, over
* the same face the roster row and tab carry, so an empty conversation still
* says whose it is.
*/
import { host, useValue, Wordmark } from '@hermes/plugin-sdk'
import { avatarColor, botAppearance, BotFace } from './avatar'
import { isBackfilledFacePng } from './avatar-image'
import { $botMeta, $lastRoster } from './data'
import { useBots } from './i18n'
import { displayName } from './labels'
import { botRosterMeta } from './routing'
import type { RosterRow } from './types'
const FACE_SIZE = 96
const FACE_GAP = 16
/** What the face costs the stack in height — the offset that puts the name on
* the center line is derived from it, so the two can never drift apart. */
const FACE_BLOCK = FACE_SIZE + FACE_GAP
/** The bot whose canonical chat this session is, if it is one. Matches the
* durable registry id or the compression-lineage tip, the same pair the
* roster's click and preview identity resolve through. */
function botForStoredId(roster: readonly RosterRow[], storedId: string): null | RosterRow {
if (!storedId || !Array.isArray(roster)) {
return null
}
return (
roster.find(bot => {
const canonical = bot?.canonical_session
return String(canonical?.id ?? '') === storedId || String(canonical?.resolved_id ?? '') === storedId
}) ?? null
)
}
/** The stored id of the chat on screen. The transcript hands its slot the
* RUNTIME id, and a canonical Bot Chat is keyed by its stored one — the same
* two id spaces that misrouted Bot Mode's RPCs in #93080. The focus store is
* the translation the rest of the plugin already trusts. */
function focusedStoredId(): string {
return String(host.state.focusedStoredSessionId?.get?.() ?? '')
}
function botForChat(roster: readonly RosterRow[], sessionId: string): null | RosterRow {
// Stored ids pass straight through, so a shell that hands us one still
// resolves; otherwise the focused chat is the empty one being looked at.
return botForStoredId(roster, sessionId) ?? botForStoredId(roster, focusedStoredId())
}
export function BotChatEmpty({ sessionId }: { sessionId: string }) {
const b = useBots()
// Subscribed, not read once: roster, metadata and focus all land after the
// transcript mounts. This is also how the state appears at all — the slot
// mounts for every empty session and only resolves to a bot once the roster
// is in hand.
const roster = useValue($lastRoster)
const allMeta = useValue($botMeta)
useValue(host.state.focusedStoredSessionId)
const bot = botForChat(roster, sessionId)
if (!bot) {
return null
}
// Route-keyed, exactly as the roster row resolves it: metadata is scoped to
// the gateway it came from, so a plain by-name read misses the entry a
// source-scoped bot's avatar actually lives under.
const meta = botRosterMeta(bot, allMeta)
const name = displayName(bot, meta)
const { color, image, shape } = botAppearance(bot.name, meta)
// Same rule the rows use: keep a real photo or pet, drop the SVG backfill so
// the math face can animate.
const photo = Boolean(image && !isBackfilledFacePng(image))
return (
<div
className="pointer-events-none flex w-full min-w-0 flex-col items-center justify-center px-0.5 py-6 text-center text-muted-foreground sm:px-6 lg:px-8"
data-slot="bot_chat_empty"
// The name reads as the title of the chat, so it — not the stack as a
// whole — is what should sit on the optical center line. Lifting the
// stack by half the face's block does exactly that: the face hangs above
// the line and the text lands on it, the same balance the splash strikes
// with nothing above its lettering.
style={{ transform: `translateY(-${FACE_BLOCK / 2}px)` }}
>
<div className="w-full min-w-0">
<div className="flex justify-center" style={{ marginBottom: FACE_GAP }}>
<BotFace
color={avatarColor(color, bot.name)}
image={photo ? image : null}
mood="idle"
name={bot.name}
shape={shape}
size={FACE_SIZE}
/>
</div>
<Wordmark className="mb-1" text={name} width="calc(80% - 1rem)" />
<p className="m-0 text-center leading-normal tracking-tight">{b.bot.chatEmpty}</p>
</div>
</div>
)
}