perf(queue): 按策略动态扩缩异步 Worker

将 River 执行容量按平台模型与用户组的有效并发策略动态调整,并为长任务续租、并发租约原子抢占和限流退避补充保护。\n\n统一平台模型限流继承语义,兼容历史 platformLimits/modelLimits,并为三个迁移图像模型建立独立并发租约。补充管理端显式继承/覆盖配置、指标、单元测试及隔离 PostgreSQL 验收。\n\n验证:go test ./... -count=1;pnpm lint;pnpm test;pnpm build;隔离 PostgreSQL 并发原子性/续租测试;128 任务与三分钟长任务动态 Worker 验收。
This commit is contained in:
2026-07-24 12:26:56 +08:00
parent 290b8c1854
commit 6c5daf29ca
37 changed files with 2365 additions and 205 deletions
+12 -1
View File
@@ -235,6 +235,8 @@ type PlatformModel struct {
Capabilities map[string]any `json:"capabilities,omitempty"`
BaseCapabilities map[string]any `json:"-"`
BaseBillingConfig map[string]any `json:"-"`
BaseRateLimitPolicy map[string]any `json:"-"`
BaseRuntimePolicySetID string `json:"-"`
BasePricingRuleSetID string `json:"-"`
PlatformPricingRuleSetID string `json:"-"`
PricingMode string `json:"pricingMode"`
@@ -245,6 +247,7 @@ type PlatformModel struct {
PermissionConfig map[string]any `json:"permissionConfig,omitempty"`
RetryPolicy map[string]any `json:"retryPolicy,omitempty"`
RateLimitPolicy map[string]any `json:"rateLimitPolicy,omitempty"`
RateLimitPolicyMode string `json:"rateLimitPolicyMode" enums:"inherit,override"`
RuntimePolicySetID string `json:"runtimePolicySetId,omitempty"`
RuntimePolicyOverride map[string]any `json:"runtimePolicyOverride,omitempty"`
CooldownUntil string `json:"cooldownUntil,omitempty"`
@@ -956,7 +959,9 @@ SELECT m.id::text, m.platform_id::text, COALESCE(m.base_model_id::text, ''), p.p
COALESCE(b.base_billing_config, '{}'::jsonb), COALESCE(b.pricing_rule_set_id::text, ''),
COALESCE(p.pricing_rule_set_id::text, ''), m.pricing_mode, COALESCE(m.discount_factor, 0)::float8,
COALESCE(m.pricing_rule_set_id::text, ''), m.billing_config_override, m.billing_config,
m.permission_config, m.retry_policy, m.rate_limit_policy, COALESCE(m.runtime_policy_set_id::text, ''), m.runtime_policy_override,
m.permission_config, m.retry_policy, m.rate_limit_policy, m.rate_limit_policy_mode,
COALESCE(m.runtime_policy_set_id::text, ''), m.runtime_policy_override,
COALESCE(b.default_rate_limit_policy, '{}'::jsonb), COALESCE(b.runtime_policy_set_id::text, ''),
COALESCE(to_char(m.cooldown_until AT TIME ZONE 'UTC', 'YYYY-MM-DD"T"HH24:MI:SS.MS"Z"'), ''),
m.enabled, m.created_at, m.updated_at
FROM platform_models m
@@ -964,6 +969,7 @@ JOIN integration_platforms p ON p.id = m.platform_id
LEFT JOIN LATERAL (
SELECT catalog.invocation_name, catalog.display_name, catalog.model_type,
catalog.capabilities, catalog.base_billing_config, catalog.pricing_rule_set_id,
catalog.default_rate_limit_policy, catalog.runtime_policy_set_id,
COALESCE((
SELECT jsonb_agg(DISTINCT compatibility_alias.alias ORDER BY compatibility_alias.alias)
FROM model_compatibility_aliases compatibility_alias
@@ -1004,6 +1010,7 @@ ORDER BY m.model_type ASC, m.model_name ASC`, args...)
var retryPolicy []byte
var rateLimitPolicy []byte
var runtimePolicyOverride []byte
var baseRateLimitPolicy []byte
var modelTypeBytes []byte
var legacyAliasesBytes []byte
if err := rows.Scan(
@@ -1032,8 +1039,11 @@ ORDER BY m.model_type ASC, m.model_name ASC`, args...)
&permissionConfig,
&retryPolicy,
&rateLimitPolicy,
&model.RateLimitPolicyMode,
&model.RuntimePolicySetID,
&runtimePolicyOverride,
&baseRateLimitPolicy,
&model.BaseRuntimePolicySetID,
&model.CooldownUntil,
&model.Enabled,
&model.CreatedAt,
@@ -1053,6 +1063,7 @@ ORDER BY m.model_type ASC, m.model_name ASC`, args...)
model.RetryPolicy = decodeObject(retryPolicy)
model.RateLimitPolicy = decodeObject(rateLimitPolicy)
model.RuntimePolicyOverride = decodeObject(runtimePolicyOverride)
model.BaseRateLimitPolicy = decodeObject(baseRateLimitPolicy)
models = append(models, model)
}
return models, rows.Err()