Make your computer a home for your AI partner. Avibe is the local-first Agent OS: Vibey, its built-in AI partner, works out of the box, and Claude Code, Codex & OpenCode run unchanged beside it — reach them from your browser, phone, or any chat app.
Python agentagent-osaiai-agentsai-partnerchatopsclaudeclaude-codecodexdevtoolsdiscord-botlark-botllmlocal-firstlong-term-memoryopencodeslack-botvibe-codingvibeywechat
import { useEffect, useRef, useState } from 'react';
import { useApi, type MessageSearchResult } from '../context/ApiContext';
export type UseMessageSearchOptions = {
// Minimum trimmed query length before a request fires. Below this the hook
// stays idle (results null, not loading) so an empty/near-empty palette
// doesn't hammer the endpoint. Defaults to 1.
minLength?: number;
// Debounce window in ms — the query must stop changing this long before a
// request is sent. Defaults to 200.
debounceMs?: number;
// Opt archived sessions into the results. Off by default — archived hits are
// read-only, so surfacing them is an explicit user choice, never sticky.
includeArchived?: boolean;
};
export type UseMessageSearchState = {
results: MessageSearchResult | null;
loading: boolean;
error: string | null;
};
// Debounced message-content search shared by the desktop command palette (P3)
// and the mobile search page (P4). Trims the query, gates on ``minLength``,
// debounces, and tags every request with a monotonic sequence so a slow
// earlier response can never clobber a newer one (out-of-order safety).
export function useMessageSearch(
query: string,
opts?: UseMessageSearchOptions,
): UseMessageSearchState {
const { connectWorkbenchEvents, searchMessages } = useApi();
const minLength = opts?.minLength ?? 1;
const debounceMs = opts?.debounceMs ?? 200;
const includeArchived = opts?.includeArchived ?? false;
const [results, setResults] = useState<MessageSearchResult | null>(null);
const [loading, setLoading] = useState(false);
const [error, setError] = useState<string | null>(null);
const [authorizationVersion, setAuthorizationVersion] = useState(0);
// Bumped on every fired request; a resolved response only commits if it is
// still the latest. Survives re-renders so stale in-flight calls are ignored.
const seqRef = useRef(0);
useEffect(() => connectWorkbenchEvents({
onAuthorizationChanged: () => {
seqRef.current += 1;
setResults(null);
setLoading(false);
setError(null);
setAuthorizationVersion((version) => version + 1);
},
}), [connectWorkbenchEvents]);
useEffect(() => {
const trimmed = query.trim();
// Below the threshold: clear to the idle state and skip the request. Bump
// the seq so any in-flight response from a longer prior query is dropped.
if (trimmed.length < minLength) {
seqRef.current += 1;
setResults(null);
setLoading(false);
setError(null);
return;
}
// Drop the PREVIOUS query's hits immediately (before the debounce/fetch) so
// the palette/page never render — and let the user select — stale rows under
// a changed query (Enter/tap could otherwise open a non-matching message).
// The loading spinner covers the gap until the new results resolve; the
// seq guard below still protects against out-of-order responses.
setResults(null);
setLoading(true);
setError(null);
const timer = window.setTimeout(() => {
const seq = (seqRef.current += 1);
searchMessages(trimmed, { includeArchived })
.then((res) => {
if (seq !== seqRef.current) return;
setResults(res);
setLoading(false);
})
.catch((err: unknown) => {
if (seq !== seqRef.current) return;
setResults(null);
setError(err instanceof Error ? err.message : String(err));
setLoading(false);
});
}, debounceMs);
// A new query (or unmount) cancels the pending debounce; bumping the seq
// also invalidates any request that already left the gate.
return () => {
window.clearTimeout(timer);
seqRef.current += 1;
};
}, [query, minLength, debounceMs, includeArchived, searchMessages, authorizationVersion]);
return { results, loading, error };
}