feat(intercept): 模型兜底审批单独计量 token 并在配置页展示
兜底审批(LLM judge)调模型判定漏网命令,其 token 消耗此前无法单独查看: usage 行要么无 worker 归属、要么被并进复用的模型 profile 桶里。 - llmrec 增加 WithWorker 上下文覆盖,reviewer 调用前标注 worker='judge'， 让审批调用的计量行可被单独筛出(复用现有 llm_usage.worker 列,无需迁移)。 - db.JudgeUsageStats 返回累计总量 + 近 N 天每日序列(WHERE worker='judge')。 - 新增 GET /api/intercept/judge/usage。 - 配置页「模型兜底审批」开关下方新增用量卡片:调用次数 / 输入 / 输出 / 缓存读写 五项统计 + 近 30 天每日消耗迷你柱图 + 刷新。
This commit is contained in:
Autumn-27 committed
1 parent
d0033724a6
commit
86729b65dc
8 files changed
+242
-1
No files matched your search
@@ -51,7 +51,53 @@ import {
|
||||
import { Card, CardContent } from "@/components/ui/card";
|
||||
import { Tabs, TabsContent, TabsList, TabsTrigger } from "@/components/ui/tabs";
|
||||
import { api } from "@/lib/api";
|
||||
import type { InterceptRule, InterceptAction, JudgeConfig, LLMProfile, Tool } from "@/lib/types";
|
||||
import type {
|
||||
InterceptAction,
|
||||
InterceptRule,
|
||||
JudgeConfig,
|
||||
JudgeDayUsage,
|
||||
JudgeUsage,
|
||||
LLMProfile,
|
||||
Tool,
|
||||
} from "@/lib/types";
|
||||
|
||||
// fmtTokens 把 token 数压成紧凑写法(1.2k / 3.4M),用于审批用量统计。
|
||||
function fmtTokens(n: number): string {
|
||||
if (n >= 1_000_000) return (n / 1_000_000).toFixed(1) + "M";
|
||||
if (n >= 1000) return (n / 1000).toFixed(1) + "k";
|
||||
return String(n);
|
||||
}
|
||||
|
||||
// JudgeStat 是一块统计数字(标签 + 数值)。
|
||||
function JudgeStat({ label, value }: { label: string; value: string }) {
|
||||
return (
|
||||
<div className="rounded-lg border bg-muted/20 px-3 py-2">
|
||||
<div className="text-[10px] text-muted-foreground">{label}</div>
|
||||
<div className="mt-0.5 text-lg font-semibold tabular-nums">{value}</div>
|
||||
</div>
|
||||
);
|
||||
}
|
||||
|
||||
// JudgeSparkbars 用纯 div 画近 N 天每日消耗(输入+输出)的迷你柱图,免图表库依赖。
|
||||
function JudgeSparkbars({ daily }: { daily: JudgeDayUsage[] }) {
|
||||
const max = Math.max(1, ...daily.map((d) => d.input_tokens + d.output_tokens));
|
||||
return (
|
||||
<div className="flex h-16 items-end gap-0.5">
|
||||
{daily.map((d) => {
|
||||
const total = d.input_tokens + d.output_tokens;
|
||||
const h = Math.max(2, Math.round((total / max) * 100));
|
||||
return (
|
||||
<div
|
||||
key={d.date}
|
||||
title={`${d.date} · ${d.calls} 次 · ${fmtTokens(total)} tokens`}
|
||||
className="min-w-[2px] flex-1 rounded-sm bg-violet-500/60 hover:bg-violet-500"
|
||||
style={{ height: `${h}%` }}
|
||||
/>
|
||||
);
|
||||
})}
|
||||
</div>
|
||||
);
|
||||
}
|
||||
|
||||
// ---- tool scope ----
|
||||
|
||||
@@ -164,6 +210,16 @@ function JudgeCard() {
|
||||
const [profiles, setProfiles] = React.useState<LLMProfile[]>([]);
|
||||
const [loading, setLoading] = React.useState(true);
|
||||
const [saving, setSaving] = React.useState(false);
|
||||
const [usage, setUsage] = React.useState<JudgeUsage | null>(null);
|
||||
|
||||
// 审批用量统计:失败不打断配置页,仅在开启时拉取。
|
||||
const loadUsage = React.useCallback(async () => {
|
||||
try {
|
||||
setUsage(await api.interceptJudgeUsage(30));
|
||||
} catch {
|
||||
// 忽略:统计不可用不应影响配置编辑
|
||||
}
|
||||
}, []);
|
||||
|
||||
const load = React.useCallback(async () => {
|
||||
setLoading(true);
|
||||
@@ -182,6 +238,11 @@ function JudgeCard() {
|
||||
load();
|
||||
}, [load]);
|
||||
|
||||
// 开启后(含初次加载把开关读为 true 时)拉取审批用量统计。
|
||||
React.useEffect(() => {
|
||||
if (cfg.enabled) loadUsage();
|
||||
}, [cfg.enabled, loadUsage]);
|
||||
|
||||
function patch(p: Partial<JudgeConfig>) {
|
||||
setCfg((c) => ({ ...c, ...p }));
|
||||
}
|
||||
@@ -237,6 +298,40 @@ function JudgeCard() {
|
||||
</div>
|
||||
</div>
|
||||
|
||||
{/* 审批 Token 用量统计(全局累计,独立于各模型配置) */}
|
||||
{cfg.enabled && usage && (
|
||||
<Card>
|
||||
<CardContent className="p-4">
|
||||
<div className="mb-3 flex items-center justify-between">
|
||||
<div>
|
||||
<p className="text-sm font-medium">审批 Token 消耗</p>
|
||||
<p className="text-xs text-muted-foreground">
|
||||
模型兜底审批累计用量,单独计量(worker=judge),不与各模型配置的统计混在一起
|
||||
</p>
|
||||
</div>
|
||||
<Button variant="ghost" size="sm" className="h-7 text-xs" onClick={loadUsage}>
|
||||
刷新
|
||||
</Button>
|
||||
</div>
|
||||
<div className="grid grid-cols-2 gap-3 sm:grid-cols-5">
|
||||
<JudgeStat label="审批调用" value={usage.calls.toLocaleString()} />
|
||||
<JudgeStat label="输入 Token" value={fmtTokens(usage.input_tokens)} />
|
||||
<JudgeStat label="输出 Token" value={fmtTokens(usage.output_tokens)} />
|
||||
<JudgeStat label="缓存读" value={fmtTokens(usage.cache_read_tokens)} />
|
||||
<JudgeStat label="缓存写" value={fmtTokens(usage.cache_write_tokens)} />
|
||||
</div>
|
||||
{usage.daily.length > 0 && (
|
||||
<div className="mt-4">
|
||||
<p className="mb-2 text-[10px] uppercase tracking-wider text-muted-foreground">
|
||||
近 30 天每日消耗(输入 + 输出)
|
||||
</p>
|
||||
<JudgeSparkbars daily={usage.daily} />
|
||||
</div>
|
||||
)}
|
||||
</CardContent>
|
||||
</Card>
|
||||
)}
|
||||
|
||||
{cfg.enabled && (
|
||||
<div className="grid gap-4 lg:grid-cols-5">
|
||||
{/* 左:提示词编辑器(直接展开,主区域) */}
|
||||
|
||||
@@ -53,6 +53,7 @@ import type {
|
||||
InterceptPending,
|
||||
InterceptRule,
|
||||
JudgeConfig,
|
||||
JudgeUsage,
|
||||
LLMPoolStatus,
|
||||
LLMProfile,
|
||||
LLMRecordDetail,
|
||||
@@ -1265,6 +1266,7 @@ export const api = {
|
||||
// ---- intercept LLM judge (模型兜底审批,全局配置) ----
|
||||
interceptGetJudgeConfig: () => get<JudgeConfig>("/intercept/judge"),
|
||||
interceptSetJudgeConfig: (cfg: JudgeConfig) => put<{ ok: boolean }>("/intercept/judge", cfg),
|
||||
interceptJudgeUsage: (days = 30) => get<JudgeUsage>(`/intercept/judge/usage?days=${days}`),
|
||||
|
||||
// ---- commands (tool execution history, any tool) ----
|
||||
commands: (params?: { task?: string; q?: string; page?: number; size?: number }) => {
|
||||
|
||||
@@ -2464,6 +2464,15 @@ function route(m: string, path: string, seg: string[], q: URLSearchParams, b: Re
|
||||
ask_timeout_action: "deny",
|
||||
};
|
||||
if (path === "/intercept/judge" && m === "PUT") return { ok: true };
|
||||
if (path === "/intercept/judge/usage" && m === "GET")
|
||||
return {
|
||||
calls: 0,
|
||||
input_tokens: 0,
|
||||
output_tokens: 0,
|
||||
cache_read_tokens: 0,
|
||||
cache_write_tokens: 0,
|
||||
daily: [],
|
||||
};
|
||||
|
||||
// ── 旁路提问(/btw)：demo 无旁路会话 ──
|
||||
// 必须显式命中：路径以 s 结尾会被下面的读兜底判成集合返回 []，items 就成了 undefined。
|
||||
|
||||
@@ -1332,6 +1332,22 @@ export interface JudgeConfig {
|
||||
ask_timeout_action: "allow" | "deny"; // 审批超时后的默认动作
|
||||
}
|
||||
|
||||
// JudgeUsage: 模型兜底审批(judge 通道)的累计 token 用量 + 近 N 天每日序列。
|
||||
export interface JudgeDayUsage {
|
||||
date: string; // YYYY-MM-DD (UTC)
|
||||
calls: number;
|
||||
input_tokens: number;
|
||||
output_tokens: number;
|
||||
}
|
||||
export interface JudgeUsage {
|
||||
calls: number;
|
||||
input_tokens: number;
|
||||
output_tokens: number;
|
||||
cache_read_tokens: number;
|
||||
cache_write_tokens: number;
|
||||
daily: JudgeDayUsage[];
|
||||
}
|
||||
|
||||
export interface InterceptApprovalFilter {
|
||||
status?: InterceptPending["status"];
|
||||
decision_source?: "rule" | "model" | "unknown";
|
||||
|
||||
Reference in new issue
Block a user