Files
PaperClipAI/ui/src/hooks/useSharedPolling.ts
T
DottaandPaperclip be1fcb2b46 Fix agent sidebar liveness churn (#9358)
## Thinking Path

> - Paperclip is the open source app people use to manage AI agents for
work.
> - The board UI includes an agents sidebar so operators can see which
agents are currently active.
> - That sidebar depends on live-run polling, heartbeat events, and
cross-tab cache sharing to stay current without overloading the API.
> - The sidebar was visually churning because agents could leave the
live section immediately after a run ended, while progress events and
cross-tab broadcasts kept forcing hot query updates.
> - This pull request stabilizes live sidebar membership and makes
shared polling broadcasts monotonic/deduplicated.
> - The benefit is a calmer operator sidebar that still reflects real
live state without flashing between stale and fresh snapshots.

## Linked Issues or Issue Description

No public GitHub issue exists for this operator-facing bug.

Bug summary:

- What happened: the agents sidebar could flash or reshuffle around
active agents while live-run and heartbeat data was updating.
- Expected behavior: active and recently-active agents should remain
visually stable, and cross-tab cache sharing should not overwrite
fresher data with older snapshots.
- Reproduction context: run Paperclip with multiple tabs or rapid
live-run/progress updates and watch the agents sidebar while agents
enter/leave live execution.
- Deployment mode: local/operator board UI.

Related PR:

- Supersedes #9357, which carried the same fixes on a branch/title/body
that were not suitable for public contribution hygiene.

## What Changed

- Restored the maintainer-only warning wording in the developer skill
guide so the existing server skill-utils CI gate passes on current
master.
- Added a 120-second linger window for streamlined sidebar agent rows so
an agent does not immediately disappear from the live section as soon as
its last run ends.
- Deferred the recent-agent fallback until there are no live or
lingering agents, while keeping the live badge tied only to
actually-live runs.
- Stopped broad live-runs/heartbeats/agents-list invalidation on every
run progress event, while preserving targeted agent-detail invalidation.
- Added producer timestamps to cross-tab shared polling result messages
so older-or-equal snapshots are dropped before `setQueryData`.
- Added per-resource broadcast dedupe/rate limiting so tabs do not
rebroadcast equivalent cached data in a loop.
- Added focused coverage for sidebar linger behavior, staggered
multi-agent linger expiry, live update invalidation scope, shared
polling timestamp handling, and cross-tab broadcast dedupe.

## Verification

Run locally on the rebased PR branch:

- `pnpm --filter @paperclipai/ui exec vitest run
src/components/SidebarAgents.test.tsx
src/context/LiveUpdatesProvider.test.ts` — 44 tests passed.
- `pnpm --filter @paperclipai/ui exec vitest run
src/lib/cross-tab-poll.test.ts src/hooks/useSharedPolling.test.ts` — 10
tests passed.
- `pnpm --filter @paperclipai/ui typecheck` — exit code 0.

## Risks

- Low migration risk: the sidebar/polling changes are UI/client cache
behavior only, with no database or API contract changes.
- Sidebar visibility now intentionally lingers for 120 seconds after the
last live run; stale rows could remain briefly visible, but their live
badge is removed when they are no longer actually live.
- Cross-tab broadcasts are now more conservative; a missed publish
should be corrected by the next normal poll or accepted newer timestamp.

> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected — check the roadmap
first. See `CONTRIBUTING.md`.

## Model Used

OpenAI GPT-5 Codex via the Paperclip Codex coding-agent runtime; exact
API model identifier and context-window size are not exposed in this
environment. The agent used terminal/tool execution for repository
inspection, focused tests, branch preparation, and PR creation.

## Checklist

- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge

---------

Co-authored-by: Paperclip <noreply@paperclip.ing>
2026-07-10 11:33:02 -05:00

149 lines
4.9 KiB
TypeScript

import { useCallback, useEffect, useMemo, useRef, useState } from "react";
import { type QueryClient, type QueryKey, useQueryClient } from "@tanstack/react-query";
import {
SharedPollingCoordinator,
type SharedMessage,
type SharedPollingSnapshot,
} from "../lib/cross-tab-poll";
type RefetchInterval = number | false;
export interface SharedPollingQueryOptions {
companyId: string | null | undefined;
resourceKey: string;
queryKey: QueryKey;
enabled?: boolean;
refetchInterval?: RefetchInterval;
/**
* When true, non-leader tabs disable their active polling/query fetch for this
* resource and render from broadcast/cache data. Use for interval-polled hot
* endpoints. Leave false for one-shot reads where blank first paint is worse
* than a duplicated initial fetch.
*/
leaderOnly?: boolean;
}
export interface SharedPollingQueryState<TData> {
isLeader: boolean;
enabled: boolean;
refetchInterval: RefetchInterval;
publish: (data: TData | undefined, dataUpdatedAt: number) => void;
}
type RegistryEntry = {
coordinator: SharedPollingCoordinator;
refs: number;
};
const coordinators = new Map<string, RegistryEntry>();
function acquireCoordinator(companyId: string): SharedPollingCoordinator {
const existing = coordinators.get(companyId);
if (existing) {
existing.refs += 1;
return existing.coordinator;
}
const coordinator = new SharedPollingCoordinator(companyId);
coordinator.start();
coordinators.set(companyId, { coordinator, refs: 1 });
return coordinator;
}
function releaseCoordinator(companyId: string): void {
const existing = coordinators.get(companyId);
if (!existing) return;
existing.refs -= 1;
if (existing.refs > 0) return;
existing.coordinator.stop();
coordinators.delete(companyId);
}
function resourceKey(companyId: string, key: string, queryKeyHash: string): string {
return `${companyId}:${key}:${queryKeyHash}`;
}
export function applySharedPollingResult<TData>(
queryClient: Pick<QueryClient, "getQueryState" | "setQueryData">,
queryKey: QueryKey,
message: SharedMessage,
): boolean {
if (message.type !== "result") return false;
const incomingUpdatedAt = message.dataUpdatedAt ?? message.at;
if (incomingUpdatedAt <= 0) return false;
const localUpdatedAt = queryClient.getQueryState(queryKey)?.dataUpdatedAt ?? 0;
if (localUpdatedAt >= incomingUpdatedAt) return false;
queryClient.setQueryData(queryKey, message.data as TData, { updatedAt: incomingUpdatedAt });
return true;
}
export function useSharedPollingQuery<TData>({
companyId,
resourceKey: rawResourceKey,
queryKey,
enabled = true,
refetchInterval = false,
leaderOnly = false,
}: SharedPollingQueryOptions): SharedPollingQueryState<TData> {
const queryClient = useQueryClient();
const activeCompanyId = enabled && companyId ? companyId : null;
const queryKeyHash = useMemo(() => JSON.stringify(queryKey), [queryKey]);
const fullResourceKey = activeCompanyId ? resourceKey(activeCompanyId, rawResourceKey, queryKeyHash) : null;
const queryKeyRef = useRef(queryKey);
const [snapshot, setSnapshot] = useState<SharedPollingSnapshot>({ isLeader: true });
useEffect(() => {
queryKeyRef.current = queryKey;
}, [queryKey, queryKeyHash]);
useEffect(() => {
if (!activeCompanyId || !fullResourceKey) {
setSnapshot({ isLeader: true });
return;
}
const coordinator = acquireCoordinator(activeCompanyId);
const unsubscribeState = coordinator.subscribe(setSnapshot);
const unsubscribeResource = coordinator.subscribeResource(fullResourceKey, (message) => {
applySharedPollingResult(queryClient, queryKeyRef.current, message);
});
coordinator.request(fullResourceKey);
return () => {
unsubscribeResource();
unsubscribeState();
releaseCoordinator(activeCompanyId);
};
}, [activeCompanyId, fullResourceKey, leaderOnly, queryClient, queryKeyHash]);
const isLeader = !leaderOnly || snapshot.isLeader;
const queryEnabled = enabled && (!leaderOnly || snapshot.isLeader);
const coordinatedInterval = leaderOnly && !snapshot.isLeader ? false : refetchInterval;
const publish = useCallback((data: TData | undefined, dataUpdatedAt: number) => {
if (!activeCompanyId || !fullResourceKey || data === undefined) return;
const entry = coordinators.get(activeCompanyId);
entry?.coordinator.publish(fullResourceKey, data, dataUpdatedAt);
}, [activeCompanyId, fullResourceKey]);
return useMemo(
() => ({
isLeader,
enabled: queryEnabled,
refetchInterval: coordinatedInterval,
publish,
}),
[coordinatedInterval, isLeader, publish, queryEnabled],
);
}
export function usePublishSharedQueryData<TData>(
shared: Pick<SharedPollingQueryState<TData>, "publish" | "isLeader">,
data: TData | undefined,
dataUpdatedAt: number,
): void {
useEffect(() => {
if (!shared.isLeader || dataUpdatedAt <= 0) return;
shared.publish(data, dataUpdatedAt);
}, [data, dataUpdatedAt, shared]);
}