refactor: 共享请求管道、按需懒加载、流式内存护栏与连接池治理

- 抽取 ApiRequestPipeline 单例,44 个路由文件共享同一实例
- Handler/Adapter 模块级 __getattr__ 延迟导入,减少启动时间
- 新增 ensure_stream_buffer_limit() 流式内存护栏(16MB 单行 / 32MB 总量)
- HTTP 空闲连接清理与 curl_cffi LRU 会话池
- ensure_providers_bootstrapped 按需引导指定 provider_types
- Usage 事件序列化迁移至 msgpack,Redis codec 隔离
- 启动预热任务(/readyz 就绪门控)与优雅关闭
- 通知邮件模块独立开关与 SMTP 配置校验
- CryptoService DCL 线程安全修复
- 通知模块开关 DB 查询 30s 内存缓存
- /readyz 对 unknown 状态返回 503
- 预热关闭 5s 超时保护
- 预热适配器逐个 try-except 容错
- FormatConversionRegistry 哨兵模式防并发重复物化
- 流式缓冲检查无条件执行

Closes #230

Co-authored-by: AAEE86 <ppk0227@hotmail.com>
This commit is contained in:
fawney19
2026-03-14 11:59:07 +08:00
parent 45985f1c04
commit e0286aebe3
111 changed files with 2775 additions and 1102 deletions

View File

@@ -19,11 +19,9 @@ def _make_key(
return key
@patch(
"src.services.scheduling.candidate_builder.health_monitor.get_circuit_breaker_status",
return_value=(True, None),
)
def test_kiro_quota_remaining_zero_skips(_mock_cb: MagicMock) -> None:
@patch("src.services.scheduling.candidate_builder.get_health_monitor")
def test_kiro_quota_remaining_zero_skips(mock_get_health_monitor: MagicMock) -> None:
mock_get_health_monitor.return_value.get_circuit_breaker_status.return_value = (True, None)
scheduler = CacheAwareScheduler()
key = _make_key(upstream_metadata={"kiro": {"remaining": 0.0}})
@@ -38,11 +36,9 @@ def test_kiro_quota_remaining_zero_skips(_mock_cb: MagicMock) -> None:
assert reason == "Kiro 账号配额剩余 0"
@patch(
"src.services.scheduling.candidate_builder.health_monitor.get_circuit_breaker_status",
return_value=(True, None),
)
def test_kiro_quota_remaining_positive_allows(_mock_cb: MagicMock) -> None:
@patch("src.services.scheduling.candidate_builder.get_health_monitor")
def test_kiro_quota_remaining_positive_allows(mock_get_health_monitor: MagicMock) -> None:
mock_get_health_monitor.return_value.get_circuit_breaker_status.return_value = (True, None)
scheduler = CacheAwareScheduler()
key = _make_key(upstream_metadata={"kiro": {"remaining": 1.0}})
@@ -57,11 +53,9 @@ def test_kiro_quota_remaining_positive_allows(_mock_cb: MagicMock) -> None:
assert reason is None
@patch(
"src.services.scheduling.candidate_builder.health_monitor.get_circuit_breaker_status",
return_value=(True, None),
)
def test_codex_weekly_quota_exhausted_skips(_mock_cb: MagicMock) -> None:
@patch("src.services.scheduling.candidate_builder.get_health_monitor")
def test_codex_weekly_quota_exhausted_skips(mock_get_health_monitor: MagicMock) -> None:
mock_get_health_monitor.return_value.get_circuit_breaker_status.return_value = (True, None)
scheduler = CacheAwareScheduler()
key = _make_key(
upstream_metadata={
@@ -83,11 +77,9 @@ def test_codex_weekly_quota_exhausted_skips(_mock_cb: MagicMock) -> None:
assert reason == "Codex 周限额剩余 0%"
@patch(
"src.services.scheduling.candidate_builder.health_monitor.get_circuit_breaker_status",
return_value=(True, None),
)
def test_codex_5h_quota_exhausted_skips(_mock_cb: MagicMock) -> None:
@patch("src.services.scheduling.candidate_builder.get_health_monitor")
def test_codex_5h_quota_exhausted_skips(mock_get_health_monitor: MagicMock) -> None:
mock_get_health_monitor.return_value.get_circuit_breaker_status.return_value = (True, None)
scheduler = CacheAwareScheduler()
key = _make_key(
upstream_metadata={
@@ -109,11 +101,9 @@ def test_codex_5h_quota_exhausted_skips(_mock_cb: MagicMock) -> None:
assert reason == "Codex 5H 限额剩余 0%"
@patch(
"src.services.scheduling.candidate_builder.health_monitor.get_circuit_breaker_status",
return_value=(True, None),
)
def test_codex_ignores_unrelated_metadata_fields(_mock_cb: MagicMock) -> None:
@patch("src.services.scheduling.candidate_builder.get_health_monitor")
def test_codex_ignores_unrelated_metadata_fields(mock_get_health_monitor: MagicMock) -> None:
mock_get_health_monitor.return_value.get_circuit_breaker_status.return_value = (True, None)
scheduler = CacheAwareScheduler()
key = _make_key(
upstream_metadata={
@@ -136,11 +126,9 @@ def test_codex_ignores_unrelated_metadata_fields(_mock_cb: MagicMock) -> None:
assert reason is None
@patch(
"src.services.scheduling.candidate_builder.health_monitor.get_circuit_breaker_status",
return_value=(True, None),
)
def test_antigravity_model_quota_exhausted_skips(_mock_cb: MagicMock) -> None:
@patch("src.services.scheduling.candidate_builder.get_health_monitor")
def test_antigravity_model_quota_exhausted_skips(mock_get_health_monitor: MagicMock) -> None:
mock_get_health_monitor.return_value.get_circuit_breaker_status.return_value = (True, None)
scheduler = CacheAwareScheduler()
key = _make_key(
upstream_metadata={
@@ -164,11 +152,9 @@ def test_antigravity_model_quota_exhausted_skips(_mock_cb: MagicMock) -> None:
assert reason == "Antigravity 模型 ag-model 配额剩余 0%"
@patch(
"src.services.scheduling.candidate_builder.health_monitor.get_circuit_breaker_status",
return_value=(True, None),
)
def test_antigravity_other_model_not_exhausted_allows(_mock_cb: MagicMock) -> None:
@patch("src.services.scheduling.candidate_builder.get_health_monitor")
def test_antigravity_other_model_not_exhausted_allows(mock_get_health_monitor: MagicMock) -> None:
mock_get_health_monitor.return_value.get_circuit_breaker_status.return_value = (True, None)
scheduler = CacheAwareScheduler()
key = _make_key(
upstream_metadata={
@@ -192,11 +178,9 @@ def test_antigravity_other_model_not_exhausted_allows(_mock_cb: MagicMock) -> No
assert reason is None
@patch(
"src.services.scheduling.candidate_builder.health_monitor.get_circuit_breaker_status",
return_value=(True, None),
)
def test_antigravity_quota_uses_mapping_matched_model(_mock_cb: MagicMock) -> None:
@patch("src.services.scheduling.candidate_builder.get_health_monitor")
def test_antigravity_quota_uses_mapping_matched_model(mock_get_health_monitor: MagicMock) -> None:
mock_get_health_monitor.return_value.get_circuit_breaker_status.return_value = (True, None)
scheduler = CacheAwareScheduler()
# Request uses GlobalModel.name, but allowed_models only contains provider-side model id.