Consider it done. The open-source AI agent that works out of the box · 想到,就能做到。开源、开箱即用的 AI Agent。
TypeScript agentai-agentai-assistantandroidclaude-codecodexelectroniosllmmacosreact-nativetypescriptwindows
import { useEffect, useMemo } from 'react';
import type { MobileCodexRateLimitsResult } from '@cindy/maker-shared/device-link-contract';
import { getDataOwnerGeneration, isDataOwnerGenerationCurrent } from '@/contexts/dataOwnerGeneration';
import { isProviderShareAgentDeviceId } from '../../shared/providerShare';
import { useDeviceLinkReconnectEpoch } from '../features/device-link/useDeviceLinkReconnectEpoch';
import { CLAUDE_QUOTA_STATE_READ, isClaudeQuotaClear } from '../../shared/claudeQuotaState';
/**
* useRemoteDeviceUsage — device-link 远程会话各计费形态的被控端用量镜像集。
*
* 与 useRemoteClaudeSubscriptionUsage(Claude 订阅,先落地)并列,覆盖其余形态,
* 装配复用 remoteDeviceUsageMirror(invoke warm-start + push 整帧替换 + owner
* 栅栏 + CHANNEL_NOT_ALLOWED TTL):
* - Codex 账号窗口:invoke 'maker:usage:account'(agentKind='codex')返回权威组合
* payload(顶层 app-server 兼容位 + appServerBuckets 桶表 + webSnapshot WHAM 槽),
* push 'usage:codex-account-changed' 同形;槽 / 桶选择复用 useAccountUsage 导出的
* 纯函数,在消费点按会话形态与模型执行(与本机同口径,绝不显示别的桶)。
* - Claude 网关配额(LiteLLM daily/monthly):invoke 同 channel(agentKind=
* 'claude-code'),push 'usage:claude-account-changed'。
* - xAI 订阅周用量:invoke 'maker:usage:xai-subscription'(被控端 dispatch 拦截
* 执行),push 'usage:xai-subscription-changed'。
* - xAI 限流头:push-only('usage:xai-rate-limit-changed';独立账号走
* 'usage:xai-provider-rate-limit-changed' { providerId, snapshot })—— 被控端无拉取端点,
* 本机 renderer 同样只有推送缓存,远程与本机同语义降级。
*/
import {
matchCodexBucketForModel,
splitCodexAccountUsagePayload,
type CodexQuotaSource,
type RateLimitSnapshot,
} from '@/hooks/useAccountUsage';
import type { ClaudeAccountUsageSnapshot } from '@/hooks/useClaudeAccountUsage';
import { createRemoteDeviceUsageMirror } from './remoteDeviceUsageMirror';
import type { XaiRateLimitSnapshot } from '../../shared/xaiRateLimit';
import type { XaiSubscriptionUsageSnapshot } from '../../shared/xaiSubscriptionUsage';
/** 被控端 readCodexAccountUsageSnapshot 的组合 payload(与本机 push 同形)。 */
export type RemoteCodexAccountUsagePayload = RateLimitSnapshot & {
webSnapshot?: RateLimitSnapshot | null;
appServerBuckets?: Record<string, RateLimitSnapshot> | null;
};
const codexAccountMirror = createRemoteDeviceUsageMirror<RemoteCodexAccountUsagePayload>({
invokeChannel: 'maker:usage:account',
invokeArgs: ['codex'],
pushChannel: 'usage:codex-account-changed',
defaultProviderId: 'openai',
providerPushChannel: 'usage:codex-provider-account-changed',
});
/** Account cards and model lists read app-server directly, just like Mobile. */
const codexRateLimitsMirror = createRemoteDeviceUsageMirror<RemoteCodexAccountUsagePayload>({
invokeChannel: 'maker:usage:codex-rate-limits',
pushChannel: null,
defaultProviderId: 'openai',
refreshIntervalMs: 60_000,
async read(deviceId, providerId = 'openai') {
const owner = getDataOwnerGeneration();
const matchesProvider = (value: { providerId?: string }) =>
value.providerId === providerId || (value.providerId === undefined && providerId === 'openai');
try {
const result = await window.electronAPI.deviceLink.invoke(
deviceId, 'maker:usage:codex-rate-limits', [providerId],
) as MobileCodexRateLimitsResult | null;
if (result && matchesProvider(result) && result.rateLimits && result.account) {
const updatedAt = Date.now();
const buckets = result.rateLimitsByLimitId ?? {
[result.rateLimits.limitId ?? 'codex']: result.rateLimits,
};
return {
// An empty authoritative table must not revive the compatibility slot.
...(Object.keys(buckets).length ? result.rateLimits : {}),
providerId,
webSnapshot: null,
planType: result.account.planType ?? result.rateLimits.planType,
updatedAt,
appServerBuckets: Object.fromEntries(Object.entries(buckets).map(([id, bucket]) => [id, {
...bucket, planType: result.account.planType ?? bucket.planType, updatedAt,
}])),
};
}
} catch {
// Old hosts or temporary read failures retain the previous snapshot below.
}
if (!isDataOwnerGenerationCurrent(owner)) return undefined;
const fallback = await window.electronAPI.deviceLink.invoke(
deviceId, 'maker:usage:account', providerId === 'openai' ? ['codex'] : ['codex', providerId],
) as (RemoteCodexAccountUsagePayload & { providerId?: string }) | null;
// A successful empty account read is authoritative (for example after logout).
if (fallback === null) return null;
if (!fallback || !matchesProvider(fallback)) return undefined;
return fallback;
},
});
const claudeAccountMirror = createRemoteDeviceUsageMirror<ClaudeAccountUsageSnapshot>({
invokeChannel: 'maker:usage:account',
invokeArgs: ['claude-code', CLAUDE_QUOTA_STATE_READ],
decodeRead: (value) => {
if (isClaudeQuotaClear(value)) return null;
if (!value || typeof value !== 'object' || !('maxBudget' in value)) return undefined;
return value as ClaudeAccountUsageSnapshot;
},
pushChannel: 'usage:claude-account-changed',
});
const xaiSubscriptionMirror = createRemoteDeviceUsageMirror<XaiSubscriptionUsageSnapshot>({
invokeChannel: 'maker:usage:xai-subscription',
pushChannel: 'usage:xai-subscription-changed',
defaultProviderId: 'xai',
providerPushChannel: 'usage:subscription-provider-account-changed',
});
const xaiRateLimitMirror = createRemoteDeviceUsageMirror<XaiRateLimitSnapshot>({
invokeChannel: null,
pushChannel: 'usage:xai-rate-limit-changed',
defaultProviderId: 'xai',
providerPushChannel: 'usage:xai-provider-rate-limit-changed',
});
/** 被控端 Codex 账号组合 payload 镜像(远程 codex / chatgpt-bridge 会话 chip 用)。 */
export function useRemoteCodexAccountUsage(
deviceId: string | null,
providerId?: string,
{ live = false, web = false }: { live?: boolean; web?: boolean } = {},
): RemoteCodexAccountUsagePayload | null {
// The WHAM slot still comes from the existing account reader. Sharing uses a
// separate permissioned quota reader; session chips retain their push mirror.
const target = live && isProviderShareAgentDeviceId(deviceId) ? null : deviceId;
const snapshot = codexAccountMirror.useMirror(live && !web ? null : target, providerId);
const realtime = codexRateLimitsMirror.useMirror(live ? target : null, providerId);
return useMemo(() => {
if (!live) return snapshot;
if (!web) return realtime;
if (!realtime && !snapshot?.webSnapshot) return null;
return { ...realtime, webSnapshot: snapshot?.webSnapshot };
}, [live, web, realtime, snapshot]);
}
export function requestRemoteCodexAccountRefresh(deviceId: string, providerId?: string): void {
codexAccountMirror.request(deviceId, providerId);
}
/** 被控端 Claude 网关配额(LiteLLM daily/monthly)镜像(远程网关形态会话 chip 用)。 */
export function useRemoteClaudeAccountUsage(
deviceId: string | null,
): ClaudeAccountUsageSnapshot | null {
const epoch = useDeviceLinkReconnectEpoch(deviceId ?? undefined);
const usage = claudeAccountMirror.useMirror(deviceId);
useEffect(() => {
if (deviceId && epoch > 0) claudeAccountMirror.request(deviceId, undefined, true);
}, [deviceId, epoch]);
return usage;
}
/** 被控端 xAI 订阅周用量镜像(远程 xai 形态会话 chip 用)。 */
export function useRemoteXaiSubscriptionUsage(
deviceId: string | null,
providerId?: string,
): XaiSubscriptionUsageSnapshot | null {
return xaiSubscriptionMirror.useMirror(deviceId, providerId);
}
export function requestRemoteXaiSubscriptionRefresh(deviceId: string, providerId?: string): void {
xaiSubscriptionMirror.request(deviceId, providerId);
}
/** 被控端 xAI 限流头镜像(push-only,tooltip 尽力显示;独立账号订阅各自的 provider 镜像)。 */
export function useRemoteXaiRateLimit(
deviceId: string | null,
providerId?: string,
): XaiRateLimitSnapshot | null {
return xaiRateLimitMirror.useMirror(deviceId, providerId);
}
/**
* 远程组合 payload → 会话形态对应的单快照(与本机 useAccountUsage 的选槽同口径):
* chatgpt/ bridge 消耗 WHAM(openai-web)槽;Codex CLI 会话消耗 app-server 槽并按
* 当前模型匹配限额桶(matchCodexBucketForModel,匹配不到宁可不显示,绝不显示别的
* 桶);桶表为空(旧被控端 / 无 app 数据)回退顶层兼容位。split 复用本机同一实现
* (含桶表 sanitize —— 远程 payload 一律按数据对待)。
*/
export function selectRemoteCodexAccountUsage(
payload: RemoteCodexAccountUsagePayload | null,
quotaSource: CodexQuotaSource,
modelId?: string | null,
): RateLimitSnapshot | null {
if (!payload) return null;
const parts = splitCodexAccountUsagePayload(payload);
if (quotaSource === 'openai-web') return parts.web ?? null;
const buckets = parts.appServerBuckets;
if (buckets && Object.keys(buckets).length > 0) {
return matchCodexBucketForModel(buckets, modelId);
}
return parts.appServer ?? null;
}
/** 供单测重置全部镜像缓存。 */
export function resetRemoteDeviceUsageMirrorsForTest(): void {
codexAccountMirror.resetForTest();
codexRateLimitsMirror.resetForTest();
claudeAccountMirror.resetForTest();
xaiSubscriptionMirror.resetForTest();
xaiRateLimitMirror.resetForTest();
}