mirror of
https://gitcode.com/JianFeeeee/ModelRouter.git
synced 2026-10-03 23:54:06 +00:00
fbdf0dea10
三个问题都来自生产实测，不是代码审阅。
## 1. 缓存命中被计费却不被统计
网关确实从上游 usage 提取了 prompt_cache_hit_tokens（审计里能看到
cache_hit_tokens: 270104 / cache_reported: true，占 prompt 的 99.9%），
costFor() 也用它给缓存段定价了 —— 但**没有任何 bucket 记录它**。
结果：一个 99.88% 命中率的网关，报表显示 prompt_tokens 却看不出其中
多少是缓存读，也无从按源/模型/key 看命中率。
每个 bucket 现在多三个字段：
cache_hit_tokens 命中数（按上游上报）
cache_fresh_tokens 未命中的 prompt
cache_reported_reqs 上游确实上报了缓存数的请求数
第三个字段是刻意的：**「零命中」与「上游根本不上报」在命中总量里完全一样**，
而它们在「缓存折扣有没有生效」这个问题上含义相反。没有它就无法区分，
只能猜。
chat.go 的 payload 之前**没有** cache_reported（审计有、插件没有），
所以任何插件侧的缓存统计都只能猜 —— 已补上。
旧 state 文件的 bucket 没有这些字段：Lua 里 nil + number 会抛错，而钩子抛错
会让**该请求完全不记账**（一个统计缺口会变成静默缺口）。add() 里做了回填。
UI 增加 fresh/cache/cache% 三列 + Cache hit rate KPI；未上报的显示 n/r 而不是 0%。
## 2. Billing 页空白：render() 引用了未定义的 s
`render(st)` 里两处 KPI 写成 `s.degraded_reqs`，ReferenceError 让整个渲染
中断，所有表格停在初始的空 innerHTML。症状是「页面加载了但什么都没有」，
而 /api/plugins/billing/state 返回 200 且有真实数据 —— 载荷完全正确，
DOM 是空的。
更糟的是 refresh() 里的 `catch (e) { /* never break the page */ }` 把错误
**静默吞掉**了：网络面板一切正常，页面什么都没有。现在 catch 会
console.error（仍然不抛，装饰性组件不该拖垮宿主页，但必须留痕）。
## 3. 侧栏图标
billing 声明 icon = "💰"，而原生 tab 全是内联 SVG（stroke: currentColor）。
emoji 尺寸不对、不跟随主题。
WebUI 增加 pluginIconHTML：插件图标可以是文本，也可以是内联 SVG。
**SVG 走严格白名单**（tag + 属性都是 allowlist，不是 denylist）——
插件是在运维者浏览器里跑的第三方代码，不能"信任插件"；但也不能直接拒绝
SVG，因为那是唯一能和原生 tab 视觉一致的方式。
用真实 Chromium 验证 12 个用例，全部挡住，包括 foreignObject 里嵌 HTML
命名空间 <img onerror> 这个经典绕过（整体丢弃，所以 img/onerror 也没了）。
★ node 里没有 DOMParser/jsdom，所以没法在单测里跑这个过滤器 —— 用正则近似
会得到一个"测试通过但浏览器里失效"的过滤器，这比没有测试更糟。
顺带修了过滤器的两个真缺陷：输出里嵌套了空 `<svg></svg>`，且 viewBox
是从包装元素读的（永远是 null）而不是插件自己的，所以任何自定义 viewBox
的图标都会丢失。
## 判据（新增 7 项，全部变异验证）
写「注入脚本能否正常执行」这个守卫时我错了四次：
1. 静态扫「已声明的名字」→ 把 HTML 字符串里的 CSS 类名（class/div/td）
全报成未定义
2. 用 CSS 选择器解析器查样式表 → 报样式表本身坏了
3. 只挂 process 的 uncaughtException → 脚本在 IIFE 里异步跑，错误是
unhandledRejection，判据对原 bug 全绿
4. 只查「有没有抛错」→ render() 开头是 `if (!st) return`，传错字段是
**静默 no-op**：不抛、不打日志、不报错，只是页面空白
最终判据是：在 node 里用 DOM stub 真跑一遍，同时要求「无异常」且
「至少写进一个容器」，并监听 console.error。变异验证：还原 s → 红；
render 收到 undefined 字段 → 红。
表头/行列数一致性也有守卫：row() 加了缓存列而表头没加时，表格会整体错位
（cache% 落到 completion 列下）—— 渲染正常、有数据、但要仔细看才发现。
## 生产验证
重启后价目表与累计账完整保留（1.17 亿 prompt tokens）。
新请求缓存统计生效：cache_hit 947,436 / cache_fresh 888，
cache_reported_reqs 7 / 395（其余来自旧 state，正是该字段存在的意义）。
真实浏览器：表格 3 行、KPI 7 项、表头 name/cost/reqs/prompt/fresh/cache/cache%/completion、
SVG 图标 currentColor 渲染、控制台无 billing 错误。391 个测试全绿。
## 另发现一个无关 bug（未修）
首页 stats 图表抛 IndexSizeError: arc 半径为负（-2），在 ui/index.html 的
paintStats 附近。属状态页图表，不在本次范围。
794 lines
30 KiB
Go
794 lines
30 KiB
Go
package lua
|
||
|
||
import (
|
||
"encoding/json"
|
||
"os"
|
||
"path/filepath"
|
||
"strings"
|
||
"testing"
|
||
)
|
||
|
||
// The billing plugin ships with the gateway, so its arithmetic is a contract:
|
||
// a wrong price silently produces wrong money. These tests drive it through the
|
||
// real hook path and check the NUMBERS, not merely that it loads.
|
||
|
||
func billingVM(t *testing.T) (*Plugins, string) {
|
||
t.Helper()
|
||
dir := filepath.Join(t.TempDir(), "adapters")
|
||
vm := NewVM(dir)
|
||
if err := vm.Start(); err != nil {
|
||
t.Fatalf("vm: %v", err)
|
||
}
|
||
t.Cleanup(vm.Stop)
|
||
pdir := filepath.Join(t.TempDir(), "plugins")
|
||
ps := NewPlugins(vm, pdir)
|
||
if err := ps.SeedBundled(); err != nil {
|
||
t.Fatalf("seed: %v", err)
|
||
}
|
||
if err := ps.LoadDir(); err != nil {
|
||
t.Fatalf("load: %v", err)
|
||
}
|
||
return ps, pdir
|
||
}
|
||
|
||
// stateOf reads the plugin's published state as a generic map.
|
||
func stateOf(t *testing.T, ps *Plugins) map[string]interface{} {
|
||
t.Helper()
|
||
raw := ps.State("billing")
|
||
if raw == nil {
|
||
t.Fatal("billing published no state")
|
||
}
|
||
b, err := json.Marshal(raw)
|
||
if err != nil {
|
||
t.Fatal(err)
|
||
}
|
||
var out map[string]interface{}
|
||
if err := json.Unmarshal(b, &out); err != nil {
|
||
t.Fatal(err)
|
||
}
|
||
return out
|
||
}
|
||
|
||
func approx(t *testing.T, name string, got, want float64) {
|
||
t.Helper()
|
||
d := got - want
|
||
if d < 0 {
|
||
d = -d
|
||
}
|
||
if d > 1e-9 {
|
||
t.Errorf("%s = %v, want %v (delta %v)", name, got, want, d)
|
||
}
|
||
}
|
||
|
||
// TestBillingZeroPricesIsSafe: with no configuration the plugin must still run
|
||
// and report volume. A nil-price crash here would take out every request.
|
||
func TestBillingZeroPricesIsSafe(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "m", "source": "s", "key": "***aaaaaa", "ok": true,
|
||
"prompt_tokens": 100, "completion_tokens": 50, "time": 1750000000000,
|
||
})
|
||
st := stateOf(t, ps)
|
||
total := st["total"].(map[string]interface{})
|
||
if total["requests"].(float64) != 1 {
|
||
t.Errorf("requests = %v, want 1", total["requests"])
|
||
}
|
||
approx(t, "cost with no prices", total["cost"].(float64), 0)
|
||
}
|
||
|
||
// TestBillingModelTokenPricing: the core case. prompt and completion are priced
|
||
// SEPARATELY, which is how providers publish and how the total must come out.
|
||
func TestBillingModelTokenPricing(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
// Setting prices must NOT disturb the (still empty) totals, which is the
|
||
// whole point of the prices/state split.
|
||
if err := ps.SetState("billing", map[string]interface{}{
|
||
"prices": map[string]interface{}{
|
||
"currency": "USD",
|
||
"models": map[string]interface{}{
|
||
"gpt-5.4": map[string]interface{}{"prompt": 1.25e-6, "completion": 1e-5},
|
||
},
|
||
},
|
||
}); err != nil {
|
||
t.Fatal(err)
|
||
}
|
||
// 1000 prompt * 1.25e-6 = 0.00125 ; 500 completion * 1e-5 = 0.005
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "gpt-5.4", "source": "up", "key": "***aaaaaa", "ok": true,
|
||
"prompt_tokens": 1000, "completion_tokens": 500, "time": 1750000000000,
|
||
})
|
||
st := stateOf(t, ps)
|
||
approx(t, "total cost", st["total"].(map[string]interface{})["cost"].(float64), 0.00625)
|
||
byModel := st["by_model"].(map[string]interface{})["gpt-5.4"].(map[string]interface{})
|
||
approx(t, "model cost", byModel["cost"].(float64), 0.00625)
|
||
if byModel["completion_tokens"].(float64) != 500 {
|
||
t.Errorf("completion_tokens = %v, want 500", byModel["completion_tokens"])
|
||
}
|
||
}
|
||
|
||
// TestBillingPerRequestAndTokenCombine: a flat fee is ADDED to the token cost,
|
||
// which is how an image model can be "tokens + fixed fee".
|
||
func TestBillingPerRequestAndTokenCombine(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
if err := ps.SetState("billing", map[string]interface{}{
|
||
"prices": map[string]interface{}{
|
||
"models": map[string]interface{}{
|
||
"kolors": map[string]interface{}{"prompt": 1e-6, "completion": 2e-6, "per_request": 0.04},
|
||
},
|
||
},
|
||
}); err != nil {
|
||
t.Fatal(err)
|
||
}
|
||
// 100*1e-6 + 50*2e-6 + 0.04 = 0.0402
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "kolors", "source": "sf", "key": "***bbbbbb", "ok": true,
|
||
"prompt_tokens": 100, "completion_tokens": 50, "time": 1750000000000,
|
||
})
|
||
st := stateOf(t, ps)
|
||
approx(t, "total", st["total"].(map[string]interface{})["cost"].(float64), 0.0402)
|
||
}
|
||
|
||
// TestBillingPrecedence: keys > models > default for token prices.
|
||
func TestBillingPrecedence(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
_ = ps.SetState("billing", map[string]interface{}{
|
||
"prices": map[string]interface{}{
|
||
"default": map[string]interface{}{"prompt": 9e-6, "completion": 9e-6},
|
||
"models": map[string]interface{}{"m": map[string]interface{}{"prompt": 2e-6, "completion": 3e-6}},
|
||
"keys": map[string]interface{}{"***cccccc": map[string]interface{}{"prompt": 1e-6, "completion": 1.5e-6}},
|
||
},
|
||
})
|
||
|
||
// No key match -> model price.
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "m", "source": "s", "key": "***other", "ok": true,
|
||
"prompt_tokens": 1000, "completion_tokens": 1000, "time": 1750000000000,
|
||
})
|
||
// Key match -> key price wins.
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "m", "source": "s", "key": "***cccccc", "ok": true,
|
||
"prompt_tokens": 1000, "completion_tokens": 1000, "time": 1750000000000,
|
||
})
|
||
st := stateOf(t, ps)
|
||
// 1000*2e-6 + 1000*3e-6 = 0.005 ; 1000*1e-6 + 1000*1.5e-6 = 0.0025
|
||
approx(t, "total (model + key)", st["total"].(map[string]interface{})["cost"].(float64), 0.0075)
|
||
|
||
// An unpriced model falls back to default.
|
||
ps2, _ := billingVM(t)
|
||
_ = ps2.SetState("billing", map[string]interface{}{
|
||
"prices": map[string]interface{}{
|
||
"default": map[string]interface{}{"prompt": 9e-6, "completion": 9e-6},
|
||
},
|
||
})
|
||
ps2.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "unknown", "source": "s", "key": "***d", "ok": true,
|
||
"prompt_tokens": 1000, "completion_tokens": 1000, "time": 1750000000000,
|
||
})
|
||
approx(t, "default fallback", stateOf(t, ps2)["total"].(map[string]interface{})["cost"].(float64), 0.018)
|
||
}
|
||
|
||
// TestBillingAggregatesEveryDimension: one request must land in all four
|
||
// rollups plus the daily bucket. A missing dimension is the kind of bug a
|
||
// dashboard hides (it just renders an empty table).
|
||
func TestBillingAggregatesEveryDimension(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
_ = ps.SetState("billing", map[string]interface{}{
|
||
"models": map[string]interface{}{"m1": map[string]interface{}{"prompt": 1e-6, "completion": 1e-6}},
|
||
})
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "m1", "source": "srcA", "key": "***key01", "ok": true,
|
||
"prompt_tokens": 100, "completion_tokens": 100, "time": 1750000000000,
|
||
})
|
||
st := stateOf(t, ps)
|
||
for _, dim := range []string{"by_source", "by_model", "by_key", "by_day"} {
|
||
m, ok := st[dim].(map[string]interface{})
|
||
if !ok || len(m) == 0 {
|
||
t.Errorf("%s is empty; a dimension is missing", dim)
|
||
}
|
||
}
|
||
if _, ok := st["by_source"].(map[string]interface{})["srcA"]; !ok {
|
||
t.Error("by_source lacks srcA")
|
||
}
|
||
if _, ok := st["by_key"].(map[string]interface{})["***key01"]; !ok {
|
||
t.Error("by_key lacks the gateway key")
|
||
}
|
||
// Milliseconds must be converted, not used as seconds: a raw 1750000000000
|
||
// would land in a year-57000 bucket.
|
||
days := st["by_day"].(map[string]interface{})
|
||
found := false
|
||
for k := range days {
|
||
if len(k) == 10 && strings.Contains(k, "-") {
|
||
found = true
|
||
}
|
||
if strings.HasPrefix(k, "5") && len(k) > 6 {
|
||
t.Errorf("by_day key %q suggests millisecond timestamps were not converted", k)
|
||
}
|
||
}
|
||
if !found {
|
||
t.Errorf("by_day has no YYYY-MM-DD key: %v", days)
|
||
}
|
||
}
|
||
|
||
// TestBillingFailedRequestPolicy: a failed request keeps its token cost (tokens
|
||
// really were consumed) but drops the flat per_request fee (never charged).
|
||
func TestBillingFailedRequestPolicy(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
_ = ps.SetState("billing", map[string]interface{}{
|
||
"prices": map[string]interface{}{
|
||
"models": map[string]interface{}{
|
||
"m": map[string]interface{}{"prompt": 1e-6, "completion": 1e-6, "per_request": 0.5},
|
||
},
|
||
},
|
||
})
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "m", "source": "s", "key": "***e", "ok": false, "status": 500,
|
||
"prompt_tokens": 1000, "completion_tokens": 0, "time": 1750000000000,
|
||
})
|
||
st := stateOf(t, ps)
|
||
// 1000*1e-6 = 0.001, flat dropped.
|
||
approx(t, "failed request", st["total"].(map[string]interface{})["cost"].(float64), 0.001)
|
||
if st["total"].(map[string]interface{})["failures"].(float64) != 1 {
|
||
t.Error("failures not counted")
|
||
}
|
||
}
|
||
|
||
// TestBillingStateAPIReplace: the admin price update must actually change
|
||
// subsequent pricing (not just be stored).
|
||
func TestBillingStateAPIReplace(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
_ = ps.SetState("billing", map[string]interface{}{
|
||
"prices": map[string]interface{}{
|
||
"models": map[string]interface{}{"m": map[string]interface{}{"prompt": 1e-6, "completion": 0}},
|
||
},
|
||
})
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "m", "source": "s", "key": "***f", "ok": true,
|
||
"prompt_tokens": 1000, "completion_tokens": 0, "time": 1750000000000,
|
||
})
|
||
approx(t, "before reprice", stateOf(t, ps)["total"].(map[string]interface{})["cost"].(float64), 0.001)
|
||
|
||
_ = ps.SetState("billing", map[string]interface{}{
|
||
"prices": map[string]interface{}{
|
||
"models": map[string]interface{}{"m": map[string]interface{}{"prompt": 2e-6, "completion": 0}},
|
||
},
|
||
})
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "m", "source": "s", "key": "***f", "ok": true,
|
||
"prompt_tokens": 1000, "completion_tokens": 0, "time": 1750000000000,
|
||
})
|
||
// 0.001 (old) + 0.002 (new price)
|
||
approx(t, "after reprice", stateOf(t, ps)["total"].(map[string]interface{})["cost"].(float64), 0.003)
|
||
}
|
||
|
||
// TestBillingPluginDeclaresUI: the shipped plugin must ship its dashboard, or
|
||
// "billing is enabled" would be true while showing the user nothing.
|
||
func TestBillingPluginDeclaresUI(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
for _, row := range ps.List() {
|
||
if row["name"] != "billing" {
|
||
continue
|
||
}
|
||
ui, ok := row["ui"].(map[string]interface{})
|
||
if !ok {
|
||
t.Fatal("billing declares no ui")
|
||
}
|
||
if page, _ := ui["page"].(string); page != "billing" {
|
||
t.Errorf("ui.page = %v, want \"billing\"", ui["page"])
|
||
}
|
||
if n, _ := ui["elements"].(int); n < 1 {
|
||
t.Error("billing contributes no element to an existing page")
|
||
}
|
||
return
|
||
}
|
||
t.Fatal("billing plugin is not loaded")
|
||
}
|
||
|
||
// TestBillingPluginLoadedByDefault: the shipped plugin must load with no
|
||
// configuration, since seeding only happens on a fresh plugin dir.
|
||
func TestBillingPluginLoadedByDefault(t *testing.T) {
|
||
ps, pdir := billingVM(t)
|
||
if ps.Count() != 1 {
|
||
t.Fatalf("expected 1 bundled plugin, got %d", ps.Count())
|
||
}
|
||
if _, err := os.Stat(filepath.Join(pdir, "billing.lua")); err != nil {
|
||
t.Errorf("billing.lua was not written to the plugin dir: %v", err)
|
||
}
|
||
}
|
||
|
||
// TestBillingCountsDegradations: the plugin must distinguish a request that had
|
||
// to drop below the top tier from one the top tier served. Without the chain
|
||
// trace these were identical in the accounts, so a quietly degraded gateway
|
||
// looked healthy while spending more per request.
|
||
func TestBillingCountsDegradations(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
_ = ps.SetState("billing", map[string]interface{}{
|
||
"prices": map[string]interface{}{
|
||
"models": map[string]interface{}{
|
||
"hi-tier": map[string]interface{}{"prompt": 1e-5, "completion": 1e-5},
|
||
"lo-tier": map[string]interface{}{"prompt": 1e-6, "completion": 1e-6},
|
||
},
|
||
},
|
||
})
|
||
|
||
// Request 1: degraded. tier 1 hard-failed, tier 2 served it.
|
||
ps.Fire(StageChainStep, map[string]interface{}{
|
||
"kind": "slot_fail", "tier": 1, "source": "t1", "model": "hi-tier",
|
||
})
|
||
ps.Fire(StageChainStep, map[string]interface{}{
|
||
"kind": "selected", "tier": 2, "source": "t2", "model": "lo-tier",
|
||
})
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "lo-tier", "source": "t2", "key": "***d1", "ok": true,
|
||
"prompt_tokens": 1000, "completion_tokens": 1000,
|
||
"degraded": true, "tier_served": 2, "time": 1750000000000,
|
||
})
|
||
|
||
// Request 2: clean, served by the top tier.
|
||
ps.Fire(StageChainStep, map[string]interface{}{
|
||
"kind": "selected", "tier": 1, "source": "t1", "model": "hi-tier",
|
||
})
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "hi-tier", "source": "t1", "key": "***d1", "ok": true,
|
||
"prompt_tokens": 1000, "completion_tokens": 1000,
|
||
"degraded": false, "tier_served": 1, "time": 1750000000000,
|
||
})
|
||
|
||
st := stateOf(t, ps)
|
||
if got := st["degraded_reqs"].(float64); got != 1 {
|
||
t.Errorf("degraded_reqs = %v, want 1 (one of the two requests dropped a tier)", got)
|
||
}
|
||
tiers := st["by_tier_served"].(map[string]interface{})
|
||
if tiers["2"].(float64) != 1 {
|
||
t.Errorf("by_tier_served[2] = %v, want 1", tiers["2"])
|
||
}
|
||
if tiers["1"].(float64) != 1 {
|
||
t.Errorf("by_tier_served[1] = %v, want 1", tiers["1"])
|
||
}
|
||
// Cost reflects the model actually served, not the one that should have been.
|
||
// 1000*1e-6*2 = 0.002 for the degraded one, 1000*1e-5*2 = 0.02 for the clean one.
|
||
approx(t, "total", st["total"].(map[string]interface{})["cost"].(float64), 0.022)
|
||
}
|
||
|
||
// TestBillingAggregatesSkipReasons: skip reasons are the actionable diagnostic
|
||
// ("no schedulable slot (cooling or quota exhausted)"), so they must be
|
||
// counted. The wait time is normalised, otherwise a fresh row per request would
|
||
// appear whenever the busy-wait text varies.
|
||
func TestBillingAggregatesSkipReasons(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
ps.Fire(StageChainStep, map[string]interface{}{
|
||
"kind": "tier_skip", "tier": 1, "reason": "no schedulable slot (cooling or quota exhausted)",
|
||
})
|
||
ps.Fire(StageChainStep, map[string]interface{}{
|
||
"kind": "tier_busy", "tier": 2, "reason": "no free slot within 2s",
|
||
})
|
||
ps.Fire(StageChainStep, map[string]interface{}{
|
||
"kind": "tier_busy", "tier": 3, "reason": "no free slot within 2.0001s",
|
||
})
|
||
st := stateOf(t, ps)
|
||
reasons := st["skip_reasons"].(map[string]interface{})
|
||
if len(reasons) != 2 {
|
||
t.Errorf("skip_reasons = %v, want 2 (the two variable waits must collapse to one)", reasons)
|
||
}
|
||
busy, ok := reasons["no free slot within <wait>"]
|
||
if !ok {
|
||
t.Errorf("busy reason missing; got %v", reasons)
|
||
} else if busy.(float64) != 2 {
|
||
t.Errorf("busy count = %v, want 2 (two different wait texts, one cause)", busy)
|
||
}
|
||
}
|
||
|
||
// ---- prompt-cache pricing -------------------------------------------------
|
||
//
|
||
// A cached prompt token is not a fresh one. Charging the full prompt rate made a
|
||
// 1M-token request of which 900k were cache reads cost 10 USD instead of ~1.9
|
||
// — an order of magnitude, on exactly the traffic the cache exists to make
|
||
// cheap. Agent traffic replays long shared prefixes constantly, so this was the
|
||
// single largest source of over-billing in the plugin.
|
||
|
||
func TestBillingCacheHitsAreDiscounted(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
_ = ps.SetState("billing", map[string]interface{}{
|
||
"prices": map[string]interface{}{
|
||
"models": map[string]interface{}{
|
||
"m": map[string]interface{}{"prompt": 1e-5, "completion": 1e-5},
|
||
},
|
||
},
|
||
})
|
||
// 1M prompt of which 900k cached, default discount 0.1
|
||
// 100k fresh * 1e-5 = 1.0 ; 900k cached * 1e-5 * 0.1 = 0.9
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "m", "source": "s", "key": "***c", "ok": true,
|
||
"prompt_tokens": 1000000, "completion_tokens": 0,
|
||
"cache_hit_tokens": 900000, "time": 1750000000000,
|
||
})
|
||
approx(t, "cache-discounted cost", stateOf(t, ps)["total"].(map[string]interface{})["cost"].(float64), 1.9)
|
||
}
|
||
|
||
// A per-model discount overrides the global one, because the ratio is a
|
||
// per-provider fact, not a constant.
|
||
func TestBillingCacheDiscountIsPerModel(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
_ = ps.SetState("billing", map[string]interface{}{
|
||
"prices": map[string]interface{}{
|
||
"models": map[string]interface{}{
|
||
"free-cache": map[string]interface{}{
|
||
"prompt": 1e-5, "completion": 0, "cache_discount": 0,
|
||
},
|
||
"flat": map[string]interface{}{
|
||
"prompt": 1e-5, "completion": 0, "cache_discount": 1,
|
||
},
|
||
},
|
||
},
|
||
})
|
||
for _, m := range []string{"free-cache", "flat"} {
|
||
ps2, _ := billingVM(t)
|
||
_ = ps2.SetState("billing", map[string]interface{}{
|
||
"prices": map[string]interface{}{
|
||
"models": map[string]interface{}{
|
||
m: map[string]interface{}{"prompt": 1e-5, "completion": 0, "cache_discount": map[bool]float64{true: 0, false: 1}[m == "free-cache"]},
|
||
},
|
||
},
|
||
})
|
||
ps2.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": m, "source": "s", "key": "***c", "ok": true,
|
||
"prompt_tokens": 1000000, "cache_hit_tokens": 1000000,
|
||
"time": 1750000000000,
|
||
})
|
||
want := 0.0
|
||
if m == "flat" {
|
||
want = 10.0
|
||
}
|
||
approx(t, m+" (all cached)", stateOf(t, ps2)["total"].(map[string]interface{})["cost"].(float64), want)
|
||
}
|
||
}
|
||
|
||
// A misbehaving adapter reporting more cache hits than prompt tokens must not
|
||
// produce negative fresh tokens.
|
||
func TestBillingCacheHitClampedToPrompt(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
_ = ps.SetState("billing", map[string]interface{}{
|
||
"prices": map[string]interface{}{
|
||
"models": map[string]interface{}{"m": map[string]interface{}{"prompt": 1e-5, "completion": 0}},
|
||
},
|
||
})
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "m", "source": "s", "key": "***c", "ok": true,
|
||
"prompt_tokens": 100, "completion_tokens": 0,
|
||
"cache_hit_tokens": 999999, // nonsense from a broken adapter
|
||
"time": 1750000000000,
|
||
})
|
||
// Clamped to 100 cached, 0 fresh => 100 * 1e-5 * 0.1
|
||
got := stateOf(t, ps)["total"].(map[string]interface{})["cost"].(float64)
|
||
if got < 0 {
|
||
t.Errorf("cost = %v, must never be negative", got)
|
||
}
|
||
approx(t, "clamped cost", got, 0.0001)
|
||
}
|
||
|
||
// ---- unpriced traffic -----------------------------------------------------
|
||
|
||
// An unpriced model silently costing 0 is the most dangerous failure a cost
|
||
// plugin has: the bill still adds up, it just quietly under-reports, and
|
||
// nothing looks broken. It must be counted and named.
|
||
func TestBillingCountsUnpricedTraffic(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
_ = ps.SetState("billing", map[string]interface{}{
|
||
"prices": map[string]interface{}{
|
||
"models": map[string]interface{}{
|
||
"priced": map[string]interface{}{"prompt": 1e-5},
|
||
},
|
||
},
|
||
})
|
||
// 100k+100k tokens on a model with no price entry.
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "MYSTERY-MODEL", "source": "s", "key": "***u", "ok": true,
|
||
"prompt_tokens": 100000, "completion_tokens": 100000, "time": 1750000000000,
|
||
})
|
||
// A priced one, to prove the counter is selective.
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "priced", "source": "s", "key": "***u", "ok": true,
|
||
"prompt_tokens": 1000, "completion_tokens": 0, "time": 1750000000000,
|
||
})
|
||
st := stateOf(t, ps)
|
||
if got := st["unpriced_reqs"].(float64); got != 1 {
|
||
t.Errorf("unpriced_reqs = %v, want 1 (only the mystery model)", got)
|
||
}
|
||
models := st["unpriced_models"].(map[string]interface{})
|
||
if models["MYSTERY-MODEL"].(float64) != 1 {
|
||
t.Errorf("unpriced_models = %v, want MYSTERY-MODEL counted", models)
|
||
}
|
||
if _, present := models["priced"]; present {
|
||
t.Error("a priced model was counted as unpriced")
|
||
}
|
||
// The traffic is still recorded: "unpriced" must not mean "invisible".
|
||
if got := st["total"].(map[string]interface{})["requests"].(float64); got != 2 {
|
||
t.Errorf("total requests = %v, want 2 (unpriced traffic is still traffic)", got)
|
||
}
|
||
}
|
||
|
||
// A source-only or key-only price counts as priced: any dimension covering the
|
||
// request is enough.
|
||
func TestBillingAnyDimensionCountsAsPriced(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
_ = ps.SetState("billing", map[string]interface{}{
|
||
"prices": map[string]interface{}{
|
||
"sources": map[string]interface{}{"flat-fee": map[string]interface{}{"per_request": 0.02}},
|
||
},
|
||
})
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "any-model", "source": "flat-fee", "key": "***p", "ok": true,
|
||
"prompt_tokens": 10, "completion_tokens": 0, "time": 1750000000000,
|
||
})
|
||
st := stateOf(t, ps)
|
||
if got := st["unpriced_reqs"].(float64); got != 0 {
|
||
t.Errorf("unpriced_reqs = %v, want 0 (the source price covers it)", got)
|
||
}
|
||
approx(t, "flat fee", st["total"].(map[string]interface{})["cost"].(float64), 0.02)
|
||
}
|
||
|
||
// ---- 峰谷 / 时段定价 ---------------------------------------------------
|
||
//
|
||
// commandcode 的 DeepSeek V4 系列就是这么定价的：非高峰 17h/天，高峰 01-04 &
|
||
// 06-10 UTC 工作日，价格恰好 2 倍。这类规则用静态价目无法表达，而算错方向是
|
||
// 静默的——不会报错，只会一直算错。
|
||
//
|
||
// 时间判据的可测性：os.date("!%H") 取 UTC 小时。测试通过选择"确定落在窗口内"
|
||
// 与"确定落在窗口外"的时段来判定，不去伪造时钟（Lua 侧没有可注入的时钟，
|
||
// 伪造反而会让测试与真实行为脱节）。
|
||
|
||
func TestBillingPeakWindowDoublesOutsidePeak(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
_ = ps.SetState("billing", map[string]interface{}{
|
||
"prices": map[string]interface{}{
|
||
"models": map[string]interface{}{
|
||
"dsv41": map[string]interface{}{
|
||
"prompt": 1.5e-7, "completion": 6e-7,
|
||
"peak": map[string]interface{}{
|
||
"multiplier": 2,
|
||
// 全部 7 天全部 24 小时 ⇒ 永远命中
|
||
"windows": []interface{}{
|
||
map[string]interface{}{
|
||
"days": []interface{}{0, 1, 2, 3, 4, 5, 6},
|
||
"hours": []interface{}{[]interface{}{0, 23}},
|
||
},
|
||
},
|
||
},
|
||
},
|
||
},
|
||
},
|
||
})
|
||
// 1000 prompt + 1000 completion，非高峰 0.00075 → 命中峰谷 ×2 = 0.0015
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "dsv41", "source": "commandcode", "key": "***pk", "ok": true,
|
||
"prompt_tokens": 1000, "completion_tokens": 1000, "time": 1750000000000,
|
||
})
|
||
got := stateOf(t, ps)["total"].(map[string]interface{})["cost"].(float64)
|
||
want := (1000*1.5e-7 + 1000*6e-7) * 2
|
||
approx(t, "always-peak cost", got, want)
|
||
}
|
||
|
||
// 一个不存在的窗口（UTC 25 点不存在）⇒ 永不命中 ⇒ 静态价。
|
||
func TestBillingPeakWindowNotHit(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
_ = ps.SetState("billing", map[string]interface{}{
|
||
"prices": map[string]interface{}{
|
||
"models": map[string]interface{}{
|
||
"dsv41": map[string]interface{}{
|
||
"prompt": 1.5e-7, "completion": 6e-7,
|
||
"peak": map[string]interface{}{
|
||
"multiplier": 2,
|
||
// 星期 = {0..6} 但小时窗写成 [99,100]：永远不可能命中
|
||
"windows": []interface{}{
|
||
map[string]interface{}{
|
||
"days": []interface{}{0, 1, 2, 3, 4, 5, 6},
|
||
"hours": []interface{}{[]interface{}{99, 100}},
|
||
},
|
||
},
|
||
},
|
||
},
|
||
},
|
||
},
|
||
})
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "dsv41", "source": "commandcode", "key": "***pk", "ok": true,
|
||
"prompt_tokens": 1000, "completion_tokens": 1000, "time": 1750000000000,
|
||
})
|
||
got := stateOf(t, ps)["total"].(map[string]interface{})["cost"].(float64)
|
||
approx(t, "never-peak cost", got, 1000*1.5e-7+1000*6e-7)
|
||
}
|
||
|
||
// 星期不匹配 ⇒ 不命中。这一条正是"用本地时区算会整体偏移"要防的东西：
|
||
// 周日按 UTC 算，用本地时区可能算成周六而错误地命中工作日窗口。
|
||
func TestBillingPeakWindowDayMismatch(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
_ = ps.SetState("billing", map[string]interface{}{
|
||
"prices": map[string]interface{}{
|
||
"models": map[string]interface{}{
|
||
"dsv41": map[string]interface{}{
|
||
"prompt": 1.5e-7, "completion": 6e-7,
|
||
"peak": map[string]interface{}{
|
||
"multiplier": 2,
|
||
// 只在"不存在的星期 7"上开窗（os.date %w 只到 0..6）
|
||
"windows": []interface{}{
|
||
map[string]interface{}{
|
||
"days": []interface{}{7},
|
||
"hours": []interface{}{[]interface{}{0, 23}},
|
||
},
|
||
},
|
||
},
|
||
},
|
||
},
|
||
},
|
||
})
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "dsv41", "source": "commandcode", "key": "***pk", "ok": true,
|
||
"prompt_tokens": 1000, "completion_tokens": 1000, "time": 1750000000000,
|
||
})
|
||
got := stateOf(t, ps)["total"].(map[string]interface{})["cost"].(float64)
|
||
if got > (1000*1.5e-7+1000*6e-7)*1.5 {
|
||
t.Errorf("cost = %v: a non-matching weekday must not trigger the peak multiplier", got)
|
||
}
|
||
}
|
||
|
||
// 没有 peak 规则的条目完全不受影响（向后兼容）。
|
||
func TestBillingNoPeakRuleIsUnaffected(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
_ = ps.SetState("billing", map[string]interface{}{
|
||
"prices": map[string]interface{}{
|
||
"models": map[string]interface{}{
|
||
"plain": map[string]interface{}{"prompt": 1e-6, "completion": 2e-6},
|
||
},
|
||
},
|
||
})
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "plain", "source": "s", "key": "***n", "ok": true,
|
||
"prompt_tokens": 1000, "completion_tokens": 1000, "time": 1750000000000,
|
||
})
|
||
approx(t, "no-peak cost", stateOf(t, ps)["total"].(map[string]interface{})["cost"].(float64), 1000*1e-6+1000*2e-6)
|
||
}
|
||
|
||
// 缓存读价【不】跟着峰谷翻倍：它是另一条上游费率，观测到的非峰谷价里已经含了
|
||
// 自己的折扣，跟着翻倍会把两个折扣叠在一起。
|
||
func TestBillingPeakDoesNotDoubleCacheRead(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
_ = ps.SetState("billing", map[string]interface{}{
|
||
"prices": map[string]interface{}{
|
||
"models": map[string]interface{}{
|
||
"dsv41": map[string]interface{}{
|
||
"prompt": 1.5e-7, "completion": 6e-7, "cache_discount": 0.02,
|
||
"peak": map[string]interface{}{
|
||
"multiplier": 2,
|
||
"windows": []interface{}{map[string]interface{}{
|
||
"days": []interface{}{0, 1, 2, 3, 4, 5, 6},
|
||
"hours": []interface{}{[]interface{}{0, 23}},
|
||
}},
|
||
},
|
||
},
|
||
},
|
||
},
|
||
})
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "dsv41", "source": "commandcode", "key": "***c", "ok": true,
|
||
"prompt_tokens": 1000000, "completion_tokens": 0,
|
||
"cache_hit_tokens": 1000000, "time": 1750000000000,
|
||
})
|
||
// 全部缓存命中 ⇒ 只按 cache 价 = prompt * 0.02，且不翻倍
|
||
got := stateOf(t, ps)["total"].(map[string]interface{})["cost"].(float64)
|
||
approx(t, "cache-only cost (not doubled)", got, 1e6*1.5e-7*0.02)
|
||
}
|
||
|
||
// TestBillingTracksCacheUsage is the guard for the gap production exposed: the
|
||
// gateway had prompt_cache_hit_tokens and costFor() priced the cache leg, but no
|
||
// bucket recorded the number. On a gateway where 99.88% of prompt tokens were
|
||
// cache reads, the report showed a prompt_tokens figure with no way to tell that
|
||
// most of it was cached.
|
||
func TestBillingTracksCacheUsage(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "m", "source": "s", "key": "k", "ok": true,
|
||
"prompt_tokens": 1000, "completion_tokens": 50,
|
||
"cache_hit_tokens": 900, "cache_reported": true,
|
||
})
|
||
// A second request from a source that does not report caching at all.
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "m2", "source": "s2", "ok": true,
|
||
"prompt_tokens": 100, "completion_tokens": 10,
|
||
})
|
||
|
||
st := ps.State("billing").(map[string]interface{})
|
||
total := st["total"].(map[string]interface{})
|
||
if total["cache_hit_tokens"] != float64(900) {
|
||
t.Errorf("total.cache_hit_tokens = %v, want 900", total["cache_hit_tokens"])
|
||
}
|
||
if total["cache_fresh_tokens"] != float64(200) {
|
||
t.Errorf("total.cache_fresh_tokens = %v, want 200 (1000-900 + 100)", total["cache_fresh_tokens"])
|
||
}
|
||
// Only the first request reported a cache number.
|
||
if total["cache_reported_reqs"] != float64(1) {
|
||
t.Errorf("★ total.cache_reported_reqs = %v, want 1 — a source that never "+
|
||
"reports cache usage must be distinguishable from one reporting zero hits",
|
||
total["cache_reported_reqs"])
|
||
}
|
||
|
||
// Per-source separation.
|
||
bySrc := st["by_source"].(map[string]interface{})
|
||
s1 := bySrc["s"].(map[string]interface{})
|
||
if s1["cache_hit_tokens"] != float64(900) {
|
||
t.Errorf("by_source[s].cache_hit_tokens = %v, want 900", s1["cache_hit_tokens"])
|
||
}
|
||
s2 := bySrc["s2"].(map[string]interface{})
|
||
if s2["cache_reported_reqs"] != float64(0) {
|
||
t.Errorf("by_source[s2].cache_reported_reqs = %v, want 0", s2["cache_reported_reqs"])
|
||
}
|
||
if s2["cache_fresh_tokens"] != float64(100) {
|
||
t.Errorf("by_source[s2].cache_fresh_tokens = %v, want 100", s2["cache_fresh_tokens"])
|
||
}
|
||
}
|
||
|
||
// TestBillingCacheBucketsSurviveOlderStateFiles: a state file written before these
|
||
// fields existed must not crash the hook. `nil + number` is an error in Lua, and
|
||
// a hook that throws stops accounting for that request entirely — which is how a
|
||
// billing gap turns into a silent one.
|
||
func TestBillingCacheBucketsSurviveOlderStateFiles(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
// Simulate a state restored from an older build: buckets without the new keys.
|
||
legacy := map[string]interface{}{
|
||
"total": map[string]interface{}{
|
||
"cost": 1.0, "requests": float64(5), "prompt_tokens": float64(500),
|
||
"completion_tokens": float64(50), "failures": float64(0),
|
||
},
|
||
"by_source": map[string]interface{}{
|
||
"legacy": map[string]interface{}{"cost": float64(0), "requests": float64(5),
|
||
"prompt_tokens": float64(500), "completion_tokens": float64(50), "failures": float64(0)},
|
||
},
|
||
"by_model": map[string]interface{}{}, "by_key": map[string]interface{}{},
|
||
"by_day": map[string]interface{}{}, "started": float64(0),
|
||
}
|
||
if err := ps.SetState("billing", legacy); err != nil {
|
||
t.Fatalf("SetState: %v", err)
|
||
}
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "m", "source": "legacy", "ok": true,
|
||
"prompt_tokens": 100, "completion_tokens": 10,
|
||
"cache_hit_tokens": 60, "cache_reported": true,
|
||
})
|
||
if len(ps.HookErrors()) != 0 {
|
||
t.Fatalf("hook error on a legacy state: %v", ps.HookErrors())
|
||
}
|
||
st := ps.State("billing").(map[string]interface{})
|
||
tot := st["total"].(map[string]interface{})
|
||
if tot["requests"] != float64(6) {
|
||
t.Errorf("requests = %v, want 6 (5 legacy + 1 new)", tot["requests"])
|
||
}
|
||
if tot["cache_hit_tokens"] != float64(60) {
|
||
t.Errorf("cache_hit_tokens = %v, want 60", tot["cache_hit_tokens"])
|
||
}
|
||
lg := st["by_source"].(map[string]interface{})["legacy"].(map[string]interface{})
|
||
if lg["cache_hit_tokens"] != float64(60) {
|
||
t.Errorf("legacy bucket cache_hit_tokens = %v, want 60", lg["cache_hit_tokens"])
|
||
}
|
||
}
|
||
|
||
// TestBillingCacheHitClampedInStats: costFor clamps the cache leg, so the
|
||
// recorded numbers must be clamped the same way. A provider that reports more
|
||
// cache hits than prompt tokens must not produce negative fresh tokens.
|
||
func TestBillingCacheHitClampedInStats(t *testing.T) {
|
||
ps, _ := billingVM(t)
|
||
ps.Fire(StageRequestEnd, map[string]interface{}{
|
||
"model": "m", "source": "s", "ok": true,
|
||
"prompt_tokens": 100, "completion_tokens": 5,
|
||
"cache_hit_tokens": 5000, "cache_reported": true,
|
||
})
|
||
st := ps.State("billing").(map[string]interface{})
|
||
tot := st["total"].(map[string]interface{})
|
||
if tot["cache_hit_tokens"] != float64(100) {
|
||
t.Errorf("★ cache_hit_tokens = %v, want 100 (clamped to prompt_tokens)",
|
||
tot["cache_hit_tokens"])
|
||
}
|
||
if tot["cache_fresh_tokens"] != float64(0) {
|
||
t.Errorf("★ cache_fresh_tokens = %v, want 0, never negative",
|
||
tot["cache_fresh_tokens"])
|
||
}
|
||
}
|