mirror of
https://github.com/Wei-Shaw/sub2api.git
synced 2026-10-15 23:22:59 +08:00
新增 /v1beta/cachedContents 的创建、列表、查询、更新有效期与删除接口。 参与账号:官方地址的 AI Studio API Key 账号、Vertex service account 账号, 以及开启"上游支持显式缓存"的自定义地址 API Key 账号。 - 网关自发缓存 ID,绑定上游资源名、持有账号与创建它的 API Key (迁移 243 gemini_cached_contents);list/get 只读本地记录 - 引用缓存的 generateContent / streamGenerateContent / countTokens 改写为上游资源名,只发往持有缓存的账号,不参与会话粘性、不换号; 拒绝重复键与 snake_case 写法;在持有账号上重试耗尽后, 上游 429 / 5xx 原样透传,401 / 403 映射为 502 - 计费:创建时一次性收取缓存 token 输入费与整个有效期的存储费; 延长按库中实际推进的时长收费,纯存储费不算一次请求;缩短与删除不退费 - 存储单价随计费模型取价,来自定价覆盖文件字段 cache_storage_cost_per_token_per_hour,未配置的模型拒绝创建; 存储费同时计入账号统计成本,渠道按 token 重算成本时叠加且不重复计入 - 扣费前校验:创建在上游建成后按确定金额校验余额 / 订阅额度 / API Key 额度并登记不封顶的在途预留,不足时删除上游缓存、只收输入费; 延长在调用上游前按增量时长校验 - 有效期上限 gateway.gemini_cached_content_max_ttl_seconds,默认 86400, 有效期向上取整到秒;创建请求保留数字原文、兼容 snake_case 元数据字段 - 上游调用与落库同客户端连接脱钩;2xx 不可用响应不换号并删除已建缓存 - 缓存不存在:AI Studio 的 403、Vertex 生成请求的 400 与缓存资源接口的 404 统一识别;原生转发(含流式)不重试、不换号、不改账号状态, 向客户端返回标准 403,不暴露上游缓存资源 ID,上游原始错误登记到 ops; PATCH 遇上游 404 移除本地记录并返回 403,DELETE 视为成功 - 自建网关上游:账号开关 explicit_cache_upstream,缓存接口发往账号自身 base_url;可选 explicit_cache_upstream_max_ttl_seconds 对齐上游 TTL 上限, 显式 TTL 超限时创建跳过该账号(全部超限返回 400),未显式时按上限创建, PATCH 取本站与账号上限的较小值;绑定缓存的请求在网关账号上收到 429 / 5xx 时不重试、不限流该账号;前端创建 / 编辑账号弹窗新增开关与上限输入框 - 用量与 ops 端点统计:带缓存 ID 的路径归一化为 /v1beta/cachedContents, 缓存操作的上游端点记为 /v1beta/cachedContents
147 lines
5.9 KiB
Go
147 lines
5.9 KiB
Go
package routes
|
|
|
|
import (
|
|
"net/http"
|
|
"net/http/httptest"
|
|
"os"
|
|
"path/filepath"
|
|
"regexp"
|
|
"sort"
|
|
"strings"
|
|
"testing"
|
|
|
|
"github.com/Wei-Shaw/sub2api/internal/handler"
|
|
"github.com/Wei-Shaw/sub2api/internal/securityaudit"
|
|
servermiddleware "github.com/Wei-Shaw/sub2api/internal/server/middleware"
|
|
"github.com/gin-gonic/gin"
|
|
"github.com/stretchr/testify/require"
|
|
)
|
|
|
|
func TestEveryGatewayPOSTRouteIsClassifiedForPromptAuditCoverage(t *testing.T) {
|
|
routeSource, err := os.ReadFile("gateway.go")
|
|
require.NoError(t, err)
|
|
pattern := regexp.MustCompile(`(?:gateway|gemini|r|codexDirect|antigravityV1|antigravityV1Beta)\.POST\("([^"]+)"`)
|
|
matches := pattern.FindAllStringSubmatch(string(routeSource), -1)
|
|
actual := map[string]struct{}{}
|
|
for _, match := range matches {
|
|
actual[match[1]] = struct{}{}
|
|
}
|
|
|
|
audited := map[string][]string{
|
|
"/messages": {"gateway_handler.go", "openai_gateway_handler.go"},
|
|
"/systemone": {"gateway_systemone.go"},
|
|
"/responses": {"gateway_handler_responses.go", "openai_gateway_handler.go"},
|
|
"/responses/*subpath": {"gateway_handler_responses.go", "openai_gateway_handler.go"},
|
|
"/chat/completions": {"gateway_handler_chat_completions.go", "openai_chat_completions.go"},
|
|
"/embeddings": {"openai_embeddings.go"},
|
|
"/alpha/search": {"openai_alpha_search.go"},
|
|
"/live": {"openai_live.go"},
|
|
"/realtime/calls": {"openai_live.go"},
|
|
"/images/generations": {"openai_images.go", "grok_media.go"},
|
|
"/images/edits": {"openai_images.go", "grok_media.go"},
|
|
"/images/generations/async": {"image_task_handler.go"},
|
|
"/images/edits/async": {"image_task_handler.go"},
|
|
"/images/batches": {"batch_image_handler.go"},
|
|
"/videos": {"grok_media.go"},
|
|
"/videos/generations": {"grok_media.go"},
|
|
"/videos/edits": {"grok_media.go"},
|
|
"/videos/extensions": {"grok_media.go"},
|
|
"/models/*modelAction": {"gemini_v1beta_handler.go"},
|
|
"/cachedContents": {"gemini_cached_contents_handler.go"},
|
|
"/tts": {"grok_audio.go"},
|
|
"/web_search": {"gateway_web_search.go"},
|
|
"/x_search": {"gateway_web_search.go"},
|
|
}
|
|
excluded := map[string]string{
|
|
"/messages/count_tokens": "tokenization only; it does not execute a model request",
|
|
"/images/batches/:id/cancel": "control-plane cancellation with no user prompt",
|
|
"/stt": "speech transcription is not a text-generation prompt",
|
|
"/custom-voices": "voice profile management has no model prompt",
|
|
}
|
|
|
|
unclassified := make([]string, 0)
|
|
for route := range actual {
|
|
if _, ok := audited[route]; ok {
|
|
continue
|
|
}
|
|
if _, ok := excluded[route]; ok {
|
|
continue
|
|
}
|
|
unclassified = append(unclassified, route)
|
|
}
|
|
sort.Strings(unclassified)
|
|
require.Empty(t, unclassified, "new gateway POST routes must be audited or explicitly classified with a no-prompt reason")
|
|
|
|
for route, files := range audited {
|
|
_, exists := actual[route]
|
|
require.Truef(t, exists, "stale prompt-audit route manifest entry %s", route)
|
|
for _, filename := range files {
|
|
source, readErr := os.ReadFile(filepath.Join("..", "..", "handler", filename))
|
|
require.NoError(t, readErr)
|
|
require.Containsf(t, string(source), "checkSecurityAudit", "%s route handler %s bypasses Coordinator", route, filename)
|
|
}
|
|
}
|
|
|
|
for route, reason := range excluded {
|
|
require.NotEmpty(t, strings.TrimSpace(reason))
|
|
_, exists := actual[route]
|
|
require.Truef(t, exists, "stale excluded route %s", route)
|
|
}
|
|
}
|
|
|
|
func TestResponsesWebSocketHasFirstAndSubsequentTurnPromptGates(t *testing.T) {
|
|
routeSource, err := os.ReadFile("gateway.go")
|
|
require.NoError(t, err)
|
|
require.GreaterOrEqual(t, strings.Count(string(routeSource), `.GET("/responses"`), 2)
|
|
handlerSource, err := os.ReadFile(filepath.Join("..", "..", "handler", "openai_gateway_handler.go"))
|
|
require.NoError(t, err)
|
|
require.Contains(t, string(handlerSource), `checkSecurityAuditStage`)
|
|
require.Contains(t, string(handlerSource), `"first_turn"`)
|
|
require.Contains(t, string(handlerSource), `"subsequent_turn"`)
|
|
wsStart := strings.Index(string(handlerSource), `func (h *OpenAIGatewayHandler) ResponsesWebSocket`)
|
|
require.NotEqual(t, -1, wsStart)
|
|
wsSource := string(handlerSource)[wsStart:]
|
|
require.Less(t,
|
|
strings.Index(wsSource, `"first_turn"`),
|
|
strings.Index(wsSource, `TryAcquireUserSlotForAPIKey`),
|
|
"the first response.create gate must precede per-request user/account slots",
|
|
)
|
|
}
|
|
|
|
func TestPromptAuditAdminRoutesRejectUnauthenticatedAndNonAdminRequests(t *testing.T) {
|
|
gin.SetMode(gin.TestMode)
|
|
router := gin.New()
|
|
handlers := &handler.Handlers{Admin: &handler.AdminHandlers{
|
|
PromptAudit: securityaudit.NewPromptAdminHandler(nil),
|
|
}}
|
|
adminAuth := servermiddleware.AdminAuthMiddleware(func(c *gin.Context) {
|
|
if c.GetHeader("Authorization") == "" {
|
|
servermiddleware.AbortWithError(c, http.StatusUnauthorized, "UNAUTHORIZED", "Authorization required")
|
|
return
|
|
}
|
|
servermiddleware.AbortWithError(c, http.StatusForbidden, "FORBIDDEN", "Admin access required")
|
|
})
|
|
auditLog := servermiddleware.AuditLogMiddleware(func(c *gin.Context) { c.Next() })
|
|
stepUp := servermiddleware.StepUpAuthMiddleware(func(c *gin.Context) { c.Next() })
|
|
RegisterAdminRoutes(router.Group("/api/v1"), handlers, adminAuth, auditLog, stepUp, nil, nil)
|
|
|
|
for _, tc := range []struct {
|
|
name string
|
|
auth string
|
|
wantStatus int
|
|
}{
|
|
{name: "unauthenticated", wantStatus: http.StatusUnauthorized},
|
|
{name: "non-admin", auth: "Bearer user-token", wantStatus: http.StatusForbidden},
|
|
} {
|
|
t.Run(tc.name, func(t *testing.T) {
|
|
recorder := httptest.NewRecorder()
|
|
request := httptest.NewRequest(http.MethodGet, "/api/v1/admin/prompt-audit/config", nil)
|
|
if tc.auth != "" {
|
|
request.Header.Set("Authorization", tc.auth)
|
|
}
|
|
router.ServeHTTP(recorder, request)
|
|
require.Equal(t, tc.wantStatus, recorder.Code)
|
|
})
|
|
}
|
|
}
|