i/Aether
1
0
Fork 0
mirror of https://github.com/fawney19/Aether.git synced 2026-09-02 17:30:23 +08:00
Code Issues Packages Projects Releases Wiki Activity

feat(pool): 引入 pro_first 调度预设、Pool 候选持久化跳过与诊断信息优化

Browse Source 
- 新增 pro_first 调度预设（Pro 优先），更新 plus_first 仅针对 Plus 计划，移除 free_team_first
- Pool 内部候选（pool_key_index 不为空）跳过 DB 持久化（available/skipped/unused 均适用）
- LRU 排序新增 catalog_lru_score 回退：runtime 无记录时使用 last_used_at_unix_secs
- 执行路径 miss 诊断消息细化为中文，按 reason 分类输出可读说明
- build_local_request_candidate_status_record 补充 extra_data 和 created_at_unix_ms 字段
- OpenAI CLI 计划构建流程补充候选评估进度跟踪与 terminal reason 设置
- 前端 PoolSchedulingDialog 增加 pro_first 预设展示，修复 LRU 默认预设检测逻辑
This commit is contained in:
fawney19
2026-04-24 13:29:05 +08:00
parent f29649e3a8
commit f3c9835759
27 changed files with 950 additions and 153 deletions
Download Patch File Download Diff File Expand all files Collapse all files

36
apps/aether-gateway/src/handlers/admin/provider/pool/config.rs
View File

@@ -9,10 +9,10 @@ const POOL_ALLOWED_SCHEDULING_PRESETS: &[&str] = &[
"load_balance",
"single_account",
"priority_first",
"free_team_first",
"free_first",
"team_first",
"plus_first",
"pro_first",
"health_first",
"latency_first",
"cost_first",
@@ -28,12 +28,12 @@ fn json_u64(value: &Value) -> Option<u64> {
fn normalize_pool_preset_mode(preset: &str, raw_mode: Option<&Value>) -> Option<String> {
match preset {
"free_team_first" | "free_first" | "team_first" | "plus_first" => {
"free_first" | "team_first" | "plus_first" | "pro_first" => {
let default_mode = match preset {
"free_team_first" => "both",
"free_first" => "free_only",
"team_first" => "team_only",
"plus_first" => "plus_only",
"pro_first" => "pro_only",
_ => unreachable!("preset covered by outer match"),
};
let normalized = raw_mode
@@ -42,12 +42,10 @@ fn normalize_pool_preset_mode(preset: &str, raw_mode: Option<&Value>) -> Option<
.filter(|value| !value.is_empty())
.map(|value| value.to_ascii_lowercase())
.filter(|value| match preset {
"free_team_first" => {
matches!(value.as_str(), "both" | "free_only" | "team_only")
}
"free_first" => value == "free_only",
"team_first" => value == "team_only",
"plus_first" => value == "plus_only",
"pro_first" => value == "pro_only",
_ => false,
})
.unwrap_or_else(|| default_mode.to_string());
@@ -426,14 +424,15 @@ mod tests {
"pool_advanced": {
"scheduling_presets": [
{"preset": "cache_affinity", "enabled": false},
{"preset": "plus_first", "enabled": true, "mode": "plus_only"}
{"preset": "plus_first", "enabled": true, "mode": "plus_only"},
{"preset": "pro_first", "enabled": true, "mode": "pro_only"}
]
}
})))
.expect("pool config should parse");
assert!(!config.lru_enabled);
assert_eq!(config.scheduling_presets.len(), 2);
assert_eq!(config.scheduling_presets.len(), 3);
assert_eq!(config.scheduling_presets[0].preset, "cache_affinity");
assert!(!config.scheduling_presets[0].enabled);
assert_eq!(config.scheduling_presets[1].preset, "plus_first");
@@ -441,17 +440,22 @@ mod tests {
config.scheduling_presets[1].mode.as_deref(),
Some("plus_only")
);
assert_eq!(config.scheduling_presets[2].preset, "pro_first");
assert_eq!(
config.scheduling_presets[2].mode.as_deref(),
Some("pro_only")
);
}
#[test]
fn parses_legacy_string_style_scheduling_presets_like_python() {
fn parses_legacy_string_style_scheduling_presets() {
let config = admin_provider_pool_config_from_config_value(Some(&json!({
"pool_advanced": {
"lru_enabled": false,
"scheduling_presets": [
"free_team_first",
"free_first",
"recent_refresh",
"free_team_first"
"free_first"
]
}
})))
@@ -461,24 +465,24 @@ mod tests {
assert_eq!(config.scheduling_presets.len(), 3);
assert_eq!(config.scheduling_presets[0].preset, "lru");
assert!(!config.scheduling_presets[0].enabled);
assert_eq!(config.scheduling_presets[1].preset, "free_team_first");
assert_eq!(config.scheduling_presets[1].preset, "free_first");
assert_eq!(config.scheduling_presets[2].preset, "recent_refresh");
}
#[test]
fn invalid_free_team_first_mode_defaults_to_both() {
fn retired_free_team_first_preset_is_rejected() {
let config = admin_provider_pool_config_from_config_value(Some(&json!({
"pool_advanced": {
"scheduling_presets": [
{"preset": "free_team_first", "enabled": true, "mode": "invalid_mode"}
{"preset": "free_team_first", "enabled": true, "mode": "team_only"}
]
}
})))
.expect("pool config should parse");
assert_eq!(config.scheduling_presets.len(), 1);
assert_eq!(config.scheduling_presets[0].preset, "free_team_first");
assert_eq!(config.scheduling_presets[0].mode.as_deref(), Some("both"));
assert_eq!(config.scheduling_presets[0].preset, "lru");
assert_eq!(config.scheduling_presets[0].mode, None);
}
#[test]

257
apps/aether-gateway/src/handlers/proxy/mod.rs
View File

@@ -62,19 +62,19 @@ use std::{collections::BTreeMap, time::Instant};
use tracing::{debug, info, warn};
const OPENAI_CHAT_LOCAL_EXECUTION_RUNTIME_MISS_DETAIL: &str =
"OpenAI chat execution runtime miss did not match a Rust execution path";
"当前 OpenAI Chat Completions 请求无法在本地执行：没有匹配到可用的执行路径";
const OPENAI_RESPONSES_LOCAL_EXECUTION_RUNTIME_MISS_DETAIL: &str =
"OpenAI responses execution runtime miss did not match a Rust execution path";
"当前 OpenAI Responses 请求无法在本地执行：没有匹配到可用的执行路径";
const OPENAI_COMPACT_LOCAL_EXECUTION_RUNTIME_MISS_DETAIL: &str =
"OpenAI compact execution runtime miss did not match a Rust execution path";
"当前 OpenAI Responses Compact 请求无法在本地执行：没有匹配到可用的执行路径";
const OPENAI_VIDEO_LOCAL_EXECUTION_RUNTIME_MISS_DETAIL: &str =
"OpenAI video execution runtime miss did not match a Rust execution path";
"当前 OpenAI Video 请求无法在本地执行：没有匹配到可用的执行路径";
const CLAUDE_MESSAGES_LOCAL_EXECUTION_RUNTIME_MISS_DETAIL: &str =
"Claude messages execution runtime miss did not match a Rust execution path";
"当前 Claude Messages 请求无法在本地执行：没有匹配到可用的执行路径";
const GEMINI_PUBLIC_LOCAL_EXECUTION_RUNTIME_MISS_DETAIL: &str =
"Gemini public execution runtime miss did not match a Rust execution path";
"当前 Gemini Public 请求无法在本地执行：没有匹配到可用的执行路径";
const GEMINI_FILES_LOCAL_EXECUTION_RUNTIME_MISS_DETAIL: &str =
"Gemini files execution runtime miss did not match a Rust execution path";
"当前 Gemini Files 请求无法在本地执行：没有匹配到可用的执行路径";
const LOCAL_ROUTE_NOT_FOUND_DETAIL: &str = "Route not found";
const LOCAL_PROXY_PASSTHROUGH_REMOVED_DETAIL: &str =
"Route matched a removed compatibility passthrough; implement it in Rust or retire the route";
@@ -1258,9 +1258,7 @@ pub(crate) async fn proxy_request(
auth_api_key_concurrency_limited,
stream_request,
)
.unwrap_or_else(|| {
"AI public execution runtime miss did not match a Rust execution path".to_string()
});
.unwrap_or_else(|| "当前 AI 请求无法在本地执行：没有匹配到可用的执行路径".to_string());
let local_execution_failure_path = if auth_api_key_concurrency_limited {
EXECUTION_PATH_LOCAL_API_KEY_CONCURRENCY_LIMITED
} else {
@@ -1424,34 +1422,237 @@ fn local_execution_runtime_miss_detail(
return Some(AUTH_API_KEY_CONCURRENCY_LIMIT_REACHED_DETAIL.to_string());
}
if let Some(detail) = local_execution_runtime_miss_model_detail(diagnostic, stream_request) {
if let Some(detail) =
local_execution_runtime_miss_diagnostic_detail(decision, diagnostic, stream_request)
{
return Some(detail);
}
local_execution_runtime_miss_route_detail(decision).map(ToOwned::to_owned)
}
fn local_execution_runtime_miss_model_detail(
fn local_execution_runtime_miss_diagnostic_detail(
decision: Option<&GatewayControlDecision>,
diagnostic: Option<&LocalExecutionRuntimeMissDiagnostic>,
stream_request: bool,
) -> Option<String> {
let diagnostic = diagnostic?;
if !matches!(
diagnostic.reason.as_str(),
"candidate_list_empty" | "all_candidates_skipped"
) {
return None;
let route_label = local_execution_runtime_miss_route_label(decision);
let request_mode = local_execution_runtime_miss_request_mode(stream_request);
match diagnostic.reason.as_str() {
"candidate_list_empty" => {
return Some(local_execution_runtime_miss_candidate_list_empty_detail(
diagnostic,
request_mode,
));
}
"all_candidates_skipped" => {
return Some(local_execution_runtime_miss_all_candidates_skipped_detail(
diagnostic,
request_mode,
));
}
"missing_auth_context" => {
return Some(format!(
"请求缺少有效的用户或 API Key 认证上下文，无法选择上游提供商（{route_label}，原因代码: missing_auth_context）"
));
}
"missing_requested_model" => {
return Some(format!(
"请求缺少 model 字段，无法选择上游提供商（{route_label}，原因代码: missing_requested_model）"
));
}
"auth_snapshot_missing" => {
return Some(format!(
"当前 API Key 的本地执行配置不存在或已过期，无法选择上游提供商（{route_label}，原因代码: auth_snapshot_missing）"
));
}
"auth_snapshot_read_failed" => {
return Some(format!(
"读取 API Key 的本地执行配置失败，无法选择上游提供商（{route_label}，原因代码: auth_snapshot_read_failed）"
));
}
"decision_input_unavailable" => {
return Some(format!(
"请求缺少本地执行所需的认证、模型或配置上下文，无法选择上游提供商（{route_label}，原因代码: decision_input_unavailable）"
));
}
"execution_runtime_candidates_exhausted" => {
return Some(format!(
"已尝试所有本地执行候选提供商，但没有任何候选成功完成请求（{route_label}，原因代码: execution_runtime_candidates_exhausted）"
));
}
"candidate_evaluation_incomplete" => {
return Some(format!(
"本地执行候选评估未完成，暂时无法为本次{request_mode}请求选择上游提供商（{route_label}，原因代码: candidate_evaluation_incomplete）"
));
}
"no_local_sync_plans" | "no_local_stream_plans" => {
return Some(format!(
"找到了候选提供商，但无法为本次{request_mode}请求构建本地执行计划。请检查端点路径、认证方式、Header/Body 规则和格式转换配置（{route_label}，原因代码: {}）",
diagnostic.reason
));
}
_ => {}
}
let requested_model = diagnostic
let reason = diagnostic.reason.trim();
if reason.is_empty() {
None
} else {
Some(format!(
"当前请求无法在本地执行：{route_label} 的执行路径未就绪（原因代码: {reason}）"
))
}
}
fn local_execution_runtime_miss_candidate_list_empty_detail(
diagnostic: &LocalExecutionRuntimeMissDiagnostic,
request_mode: &str,
) -> String {
if let Some(requested_model) = diagnostic_requested_model(diagnostic) {
return format!(
"没有可用提供商支持模型 {requested_model} 的{request_mode}请求。请检查模型映射、端点启用状态和 API Key 权限（原因代码: candidate_list_empty）"
);
}
format!(
"没有可用提供商支持本次{request_mode}请求。请检查模型字段、模型映射、端点启用状态和 API Key 权限（原因代码: candidate_list_empty）"
)
}
fn local_execution_runtime_miss_all_candidates_skipped_detail(
diagnostic: &LocalExecutionRuntimeMissDiagnostic,
request_mode: &str,
) -> String {
let candidate_count = diagnostic.candidate_count.unwrap_or(0);
let skipped_count = diagnostic
.skipped_candidate_count
.unwrap_or(candidate_count);
let skipped_summary =
local_execution_runtime_miss_skip_reasons_summary(&diagnostic.skip_reasons);
let requested_model = diagnostic_requested_model(diagnostic);
match (candidate_count, skipped_summary, requested_model) {
(count, Some(summary), Some(model)) if count > 0 => format!(
"找到 {count} 个支持模型 {model} 的候选提供商，但本次{request_mode}请求全部不可用：{summary}（原因代码: all_candidates_skipped）"
),
(count, Some(summary), None) if count > 0 => format!(
"找到 {count} 个候选提供商，但本次{request_mode}请求全部不可用：{summary}（原因代码: all_candidates_skipped）"
),
(_, Some(summary), Some(model)) => format!(
"支持模型 {model} 的候选提供商全部不可用：{summary}（原因代码: all_candidates_skipped）"
),
(_, Some(summary), None) => format!(
"候选提供商全部不可用：{summary}（原因代码: all_candidates_skipped）"
),
(count, None, Some(model)) if count > 0 => format!(
"找到 {count} 个支持模型 {model} 的候选提供商，但都不满足本次{request_mode}请求要求（原因代码: all_candidates_skipped）"
),
(count, None, None) if count > 0 => format!(
"找到 {count} 个候选提供商，但都不满足本次{request_mode}请求要求（原因代码: all_candidates_skipped）"
),
(_, None, Some(model)) if skipped_count > 0 => format!(
"支持模型 {model} 的 {skipped_count} 个候选提供商都不满足本次{request_mode}请求要求（原因代码: all_candidates_skipped）"
),
_ => format!(
"候选提供商都不满足本次{request_mode}请求要求（原因代码: all_candidates_skipped）"
),
}
}
fn diagnostic_requested_model(diagnostic: &LocalExecutionRuntimeMissDiagnostic) -> Option<&str> {
diagnostic
.requested_model
.as_deref()
.map(str::trim)
.filter(|value| !value.is_empty())?;
let request_mode = if stream_request { "流式" } else { "同步" };
Some(format!(
"没有可用的提供商支持模型 {requested_model} 的{request_mode}请求"
))
.filter(|value| !value.is_empty())
}
fn local_execution_runtime_miss_skip_reasons_summary(
skip_reasons: &BTreeMap<String, usize>,
) -> Option<String> {
if skip_reasons.is_empty() {
return None;
}
Some(
skip_reasons
.iter()
.map(|(reason, count)| {
format!(
"{} {} 次",
local_execution_runtime_miss_skip_reason_label(reason),
count
)
})
.collect::<Vec<_>>()
.join("，"),
)
}
fn local_execution_runtime_miss_skip_reason_label(reason: &str) -> &str {
match reason {
"api_key_concurrency_limit_reached" => "API Key 并发已达上限",
"auth_snapshot_missing" => "API Key 本地执行配置缺失",
"endpoint_api_format_changed" => "端点 API 格式已变更",
"endpoint_inactive" => "端点未启用",
"key_api_format_disabled" => "API Key 未启用该 API 格式",
"key_inactive" => "API Key 未启用",
"key_model_disabled" => "API Key 未允许该模型",
"mapped_model_missing" => "模型映射缺失",
"provider_inactive" => "提供商未启用",
"provider_request_body_missing" => "无法构建上游请求体",
"transport_api_format_mismatch" => "传输层 API 格式不匹配",
"transport_api_format_unsupported" => "传输层不支持该 API 格式",
"transport_auth_unavailable" => "上游认证信息不可用",
"transport_body_rules_unsupported" => "Body 规则不支持本地执行",
"transport_custom_path_unsupported" => "自定义路径不支持本地执行",
"transport_header_rules_unsupported" => "Header 规则不支持本地执行",
"transport_header_rules_apply_failed" => "Header 规则应用失败",
"transport_oauth_resolution_unsupported" => "OAuth 认证解析不支持本地执行",
"transport_provider_type_unsupported" => "提供商类型不支持本地执行",
"transport_proxy_or_tls_unsupported" => "代理或 TLS 配置不支持本地执行",
"transport_proxy_unsupported" => "代理配置不支持本地执行",
"transport_snapshot_missing" => "提供商传输配置缺失",
"transport_tls_profile_unsupported" => "TLS 指纹配置不支持本地执行",
"transport_unsupported" => "传输配置不支持本地执行",
"upstream_url_missing" => "无法构建上游请求地址",
other => other,
}
}
fn local_execution_runtime_miss_request_mode(stream_request: bool) -> &'static str {
if stream_request {
"流式"
} else {
"同步"
}
}
fn local_execution_runtime_miss_route_label(
decision: Option<&GatewayControlDecision>,
) -> &'static str {
let Some(decision) = decision else {
return "AI 请求";
};
match decision.public_path.as_str() {
"/v1/chat/completions" => "OpenAI Chat Completions",
"/v1/responses" => "OpenAI Responses",
"/v1/responses/compact" => "OpenAI Responses Compact",
"/v1/messages" => "Claude Messages",
path if path.starts_with("/v1/videos") => "OpenAI Video",
path if path.starts_with("/upload/v1beta/files") || path.starts_with("/v1beta/files") => {
"Gemini Files"
}
path if decision.route_family.as_deref() == Some("gemini")
&& (path.starts_with("/v1beta/models/") || path.starts_with("/v1/models/")) =>
{
"Gemini Public"
}
_ => "AI 请求",
}
}
fn diagnostic_is_auth_api_key_concurrency_limited(
@@ -1526,12 +1727,14 @@ mod tests {
assert_eq!(
detail.as_deref(),
Some("没有可用的提供商支持模型 gpt-5.4 的流式请求")
Some(
"没有可用提供商支持模型 gpt-5.4 的流式请求。请检查模型映射、端点启用状态和 API Key 权限（原因代码: candidate_list_empty）"
)
);
}
#[test]
fn runtime_miss_detail_falls_back_to_route_default_when_reason_is_not_model_unavailable() {
fn runtime_miss_detail_returns_auth_context_message_when_auth_context_is_missing() {
let decision = GatewayControlDecision::synthetic(
"/v1/messages",
Some("ai_public".to_string()),
@@ -1550,7 +1753,9 @@ mod tests {
assert_eq!(
detail.as_deref(),
Some("Claude messages execution runtime miss did not match a Rust execution path")
Some(
"请求缺少有效的用户或 API Key 认证上下文，无法选择上游提供商（Claude Messages，原因代码: missing_auth_context）"
)
);
}