oadmin/sub2api
1
0
Fork 0
Code Issues Pull Requests 1 Actions Packages Projects Releases Wiki Activity

refactor: replace scope-level rate limiting with model-level rate limiting

Browse Source 
Merge functional changes from develop branch:
- Remove AntigravityQuotaScope system (claude/gemini_text/gemini_image)
- Replace with per-model rate limiting using resolveAntigravityModelKey
- Remove model load statistics (IncrModelCallCount/GetModelLoadBatch)
- Simplify account selection to unified priority→load→LRU algorithm
- Remove SetAntigravityQuotaScopeLimit from AccountRepository
- Clean up scope-related UI indicators and API fields
This commit is contained in:
erio
2026-02-09 08:19:01 +08:00
parent 1af06aed96
commit fc095bf054
23 changed files with 137 additions and 1162 deletions
Download Patch File Download Diff File Expand all files Collapse all files

47
backend/internal/repository/account_repo.go
View File

@@ -798,53 +798,6 @@ func (r *accountRepository) SetRateLimited(ctx context.Context, id int64, resetA
return nil
}
func (r *accountRepository) SetAntigravityQuotaScopeLimit(ctx context.Context, id int64, scope service.AntigravityQuotaScope, resetAt time.Time) error {
now := time.Now().UTC()
payload := map[string]string{
"rate_limited_at": now.Format(time.RFC3339),
"rate_limit_reset_at": resetAt.UTC().Format(time.RFC3339),
}
raw, err := json.Marshal(payload)
if err != nil {
return err
}
scopeKey := string(scope)
client := clientFromContext(ctx, r.client)
result, err := client.ExecContext(
ctx,
`UPDATE accounts SET
extra = jsonb_set(
jsonb_set(COALESCE(extra, '{}'::jsonb), '{antigravity_quota_scopes}'::text[], COALESCE(extra->'antigravity_quota_scopes', '{}'::jsonb), true),
ARRAY['antigravity_quota_scopes', $1]::text[],
$2::jsonb,
true
),
updated_at = NOW(),
last_used_at = NOW()
WHERE id = $3 AND deleted_at IS NULL`,
scopeKey,
raw,
id,
)
if err != nil {
return err
}
affected, err := result.RowsAffected()
if err != nil {
return err
}
if affected == 0 {
return service.ErrAccountNotFound
}
if err := enqueueSchedulerOutbox(ctx, r.sql, service.SchedulerOutboxEventAccountChanged, &id, nil, nil); err != nil {
log.Printf("[SchedulerOutbox] enqueue quota scope failed: account=%d err=%v", id, err)
}
return nil
}
func (r *accountRepository) SetModelRateLimit(ctx context.Context, id int64, scope string, resetAt time.Time) error {
if scope == "" {
return nil

97
backend/internal/repository/gateway_cache.go
View File

@@ -11,14 +11,6 @@ import (
const stickySessionPrefix = "sticky_session:"
// 模型负载统计相关常量
const (
modelLoadKeyPrefix = "ag:model_load:" // 模型调用次数 key 前缀
modelLastUsedKeyPrefix = "ag:model_last_used:" // 模型最后调度时间 key 前缀
modelLoadTTL = 24 * time.Hour // 调用次数 TTL（24 小时无调用后清零）
modelLastUsedTTL = 24 * time.Hour // 最后调度时间 TTL
)
type gatewayCache struct {
rdb *redis.Client
}
@@ -59,92 +51,3 @@ func (c *gatewayCache) DeleteSessionAccountID(ctx context.Context, groupID int64
key := buildSessionKey(groupID, sessionHash)
return c.rdb.Del(ctx, key).Err()
}
// ============ Antigravity 模型负载统计方法 ============
// modelLoadKey 构建模型调用次数 key
// 格式: ag:model_load:{accountID}:{model}
func modelLoadKey(accountID int64, model string) string {
return fmt.Sprintf("%s%d:%s", modelLoadKeyPrefix, accountID, model)
}
// modelLastUsedKey 构建模型最后调度时间 key
// 格式: ag:model_last_used:{accountID}:{model}
func modelLastUsedKey(accountID int64, model string) string {
return fmt.Sprintf("%s%d:%s", modelLastUsedKeyPrefix, accountID, model)
}
// IncrModelCallCount 增加模型调用次数并更新最后调度时间
// 返回更新后的调用次数
func (c *gatewayCache) IncrModelCallCount(ctx context.Context, accountID int64, model string) (int64, error) {
loadKey := modelLoadKey(accountID, model)
lastUsedKey := modelLastUsedKey(accountID, model)
pipe := c.rdb.Pipeline()
incrCmd := pipe.Incr(ctx, loadKey)
pipe.Expire(ctx, loadKey, modelLoadTTL) // 每次调用刷新 TTL
pipe.Set(ctx, lastUsedKey, time.Now().Unix(), modelLastUsedTTL)
if _, err := pipe.Exec(ctx); err != nil {
return 0, err
}
return incrCmd.Val(), nil
}
// GetModelLoadBatch 批量获取账号的模型负载信息
func (c *gatewayCache) GetModelLoadBatch(ctx context.Context, accountIDs []int64, model string) (map[int64]*service.ModelLoadInfo, error) {
if len(accountIDs) == 0 {
return make(map[int64]*service.ModelLoadInfo), nil
}
loadCmds, lastUsedCmds := c.pipelineModelLoadGet(ctx, accountIDs, model)
return c.parseModelLoadResults(accountIDs, loadCmds, lastUsedCmds), nil
}
// pipelineModelLoadGet 批量获取模型负载的 Pipeline 操作
func (c *gatewayCache) pipelineModelLoadGet(
ctx context.Context,
accountIDs []int64,
model string,
) (map[int64]*redis.StringCmd, map[int64]*redis.StringCmd) {
pipe := c.rdb.Pipeline()
loadCmds := make(map[int64]*redis.StringCmd, len(accountIDs))
lastUsedCmds := make(map[int64]*redis.StringCmd, len(accountIDs))
for _, id := range accountIDs {
loadCmds[id] = pipe.Get(ctx, modelLoadKey(id, model))
lastUsedCmds[id] = pipe.Get(ctx, modelLastUsedKey(id, model))
}
_, _ = pipe.Exec(ctx) // 忽略错误，key 不存在是正常的
return loadCmds, lastUsedCmds
}
// parseModelLoadResults 解析 Pipeline 结果
func (c *gatewayCache) parseModelLoadResults(
accountIDs []int64,
loadCmds map[int64]*redis.StringCmd,
lastUsedCmds map[int64]*redis.StringCmd,
) map[int64]*service.ModelLoadInfo {
result := make(map[int64]*service.ModelLoadInfo, len(accountIDs))
for _, id := range accountIDs {
result[id] = &service.ModelLoadInfo{
CallCount: getInt64OrZero(loadCmds[id]),
LastUsedAt: getTimeOrZero(lastUsedCmds[id]),
}
}
return result
}
// getInt64OrZero 从 StringCmd 获取 int64 值，失败返回 0
func getInt64OrZero(cmd *redis.StringCmd) int64 {
val, _ := cmd.Int64()
return val
}
// getTimeOrZero 从 StringCmd 获取 time.Time，失败返回零值
func getTimeOrZero(cmd *redis.StringCmd) time.Time {
val, err := cmd.Int64()
if err != nil {
return time.Time{}
}
return time.Unix(val, 0)
}

234
backend/internal/repository/gateway_cache_model_load_integration_test.go
View File

@@ -1,234 +0,0 @@
//go:build integration
package repository
import (
"context"
"testing"
"time"
"github.com/stretchr/testify/require"
"github.com/stretchr/testify/suite"
)
// ============ Gateway Cache 模型负载统计集成测试 ============
type GatewayCacheModelLoadSuite struct {
suite.Suite
}
func TestGatewayCacheModelLoadSuite(t *testing.T) {
suite.Run(t, new(GatewayCacheModelLoadSuite))
}
func (s *GatewayCacheModelLoadSuite) TestIncrModelCallCount_Basic() {
t := s.T()
rdb := testRedis(t)
cache := &gatewayCache{rdb: rdb}
ctx := context.Background()
accountID := int64(123)
model := "claude-sonnet-4-20250514"
// 首次调用应返回 1
count1, err := cache.IncrModelCallCount(ctx, accountID, model)
require.NoError(t, err)
require.Equal(t, int64(1), count1)
// 第二次调用应返回 2
count2, err := cache.IncrModelCallCount(ctx, accountID, model)
require.NoError(t, err)
require.Equal(t, int64(2), count2)
// 第三次调用应返回 3
count3, err := cache.IncrModelCallCount(ctx, accountID, model)
require.NoError(t, err)
require.Equal(t, int64(3), count3)
}
func (s *GatewayCacheModelLoadSuite) TestIncrModelCallCount_DifferentModels() {
t := s.T()
rdb := testRedis(t)
cache := &gatewayCache{rdb: rdb}
ctx := context.Background()
accountID := int64(456)
model1 := "claude-sonnet-4-20250514"
model2 := "claude-opus-4-5-20251101"
// 不同模型应该独立计数
count1, err := cache.IncrModelCallCount(ctx, accountID, model1)
require.NoError(t, err)
require.Equal(t, int64(1), count1)
count2, err := cache.IncrModelCallCount(ctx, accountID, model2)
require.NoError(t, err)
require.Equal(t, int64(1), count2)
count1Again, err := cache.IncrModelCallCount(ctx, accountID, model1)
require.NoError(t, err)
require.Equal(t, int64(2), count1Again)
}
func (s *GatewayCacheModelLoadSuite) TestIncrModelCallCount_DifferentAccounts() {
t := s.T()
rdb := testRedis(t)
cache := &gatewayCache{rdb: rdb}
ctx := context.Background()
account1 := int64(111)
account2 := int64(222)
model := "gemini-2.5-pro"
// 不同账号应该独立计数
count1, err := cache.IncrModelCallCount(ctx, account1, model)
require.NoError(t, err)
require.Equal(t, int64(1), count1)
count2, err := cache.IncrModelCallCount(ctx, account2, model)
require.NoError(t, err)
require.Equal(t, int64(1), count2)
}
func (s *GatewayCacheModelLoadSuite) TestGetModelLoadBatch_Empty() {
t := s.T()
rdb := testRedis(t)
cache := &gatewayCache{rdb: rdb}
ctx := context.Background()
result, err := cache.GetModelLoadBatch(ctx, []int64{}, "any-model")
require.NoError(t, err)
require.NotNil(t, result)
require.Empty(t, result)
}
func (s *GatewayCacheModelLoadSuite) TestGetModelLoadBatch_NonExistent() {
t := s.T()
rdb := testRedis(t)
cache := &gatewayCache{rdb: rdb}
ctx := context.Background()
// 查询不存在的账号应返回零值
result, err := cache.GetModelLoadBatch(ctx, []int64{9999, 9998}, "claude-sonnet-4-20250514")
require.NoError(t, err)
require.Len(t, result, 2)
require.Equal(t, int64(0), result[9999].CallCount)
require.True(t, result[9999].LastUsedAt.IsZero())
require.Equal(t, int64(0), result[9998].CallCount)
require.True(t, result[9998].LastUsedAt.IsZero())
}
func (s *GatewayCacheModelLoadSuite) TestGetModelLoadBatch_AfterIncrement() {
t := s.T()
rdb := testRedis(t)
cache := &gatewayCache{rdb: rdb}
ctx := context.Background()
accountID := int64(789)
model := "claude-sonnet-4-20250514"
// 先增加调用次数
beforeIncr := time.Now()
_, err := cache.IncrModelCallCount(ctx, accountID, model)
require.NoError(t, err)
_, err = cache.IncrModelCallCount(ctx, accountID, model)
require.NoError(t, err)
_, err = cache.IncrModelCallCount(ctx, accountID, model)
require.NoError(t, err)
afterIncr := time.Now()
// 获取负载信息
result, err := cache.GetModelLoadBatch(ctx, []int64{accountID}, model)
require.NoError(t, err)
require.Len(t, result, 1)
loadInfo := result[accountID]
require.NotNil(t, loadInfo)
require.Equal(t, int64(3), loadInfo.CallCount)
require.False(t, loadInfo.LastUsedAt.IsZero())
// LastUsedAt 应该在 beforeIncr 和 afterIncr 之间
require.True(t, loadInfo.LastUsedAt.After(beforeIncr.Add(-time.Second)) || loadInfo.LastUsedAt.Equal(beforeIncr))
require.True(t, loadInfo.LastUsedAt.Before(afterIncr.Add(time.Second)) || loadInfo.LastUsedAt.Equal(afterIncr))
}
func (s *GatewayCacheModelLoadSuite) TestGetModelLoadBatch_MultipleAccounts() {
t := s.T()
rdb := testRedis(t)
cache := &gatewayCache{rdb: rdb}
ctx := context.Background()
model := "claude-opus-4-5-20251101"
account1 := int64(1001)
account2 := int64(1002)
account3 := int64(1003) // 不调用
// account1 调用 2 次
_, err := cache.IncrModelCallCount(ctx, account1, model)
require.NoError(t, err)
_, err = cache.IncrModelCallCount(ctx, account1, model)
require.NoError(t, err)
// account2 调用 5 次
for i := 0; i < 5; i++ {
_, err = cache.IncrModelCallCount(ctx, account2, model)
require.NoError(t, err)
}
// 批量获取
result, err := cache.GetModelLoadBatch(ctx, []int64{account1, account2, account3}, model)
require.NoError(t, err)
require.Len(t, result, 3)
require.Equal(t, int64(2), result[account1].CallCount)
require.False(t, result[account1].LastUsedAt.IsZero())
require.Equal(t, int64(5), result[account2].CallCount)
require.False(t, result[account2].LastUsedAt.IsZero())
require.Equal(t, int64(0), result[account3].CallCount)
require.True(t, result[account3].LastUsedAt.IsZero())
}
func (s *GatewayCacheModelLoadSuite) TestGetModelLoadBatch_ModelIsolation() {
t := s.T()
rdb := testRedis(t)
cache := &gatewayCache{rdb: rdb}
ctx := context.Background()
accountID := int64(2001)
model1 := "claude-sonnet-4-20250514"
model2 := "gemini-2.5-pro"
// 对 model1 调用 3 次
for i := 0; i < 3; i++ {
_, err := cache.IncrModelCallCount(ctx, accountID, model1)
require.NoError(t, err)
}
// 获取 model1 的负载
result1, err := cache.GetModelLoadBatch(ctx, []int64{accountID}, model1)
require.NoError(t, err)
require.Equal(t, int64(3), result1[accountID].CallCount)
// 获取 model2 的负载（应该为 0）
result2, err := cache.GetModelLoadBatch(ctx, []int64{accountID}, model2)
require.NoError(t, err)
require.Equal(t, int64(0), result2[accountID].CallCount)
}
// ============ 辅助函数测试 ============
func (s *GatewayCacheModelLoadSuite) TestModelLoadKey_Format() {
t := s.T()
key := modelLoadKey(123, "claude-sonnet-4")
require.Equal(t, "ag:model_load:123:claude-sonnet-4", key)
}
func (s *GatewayCacheModelLoadSuite) TestModelLastUsedKey_Format() {
t := s.T()
key := modelLastUsedKey(456, "gemini-2.5-pro")
require.Equal(t, "ag:model_last_used:456:gemini-2.5-pro", key)
}