feat: delegate 协议改为 header 元数据 + gzip 压缩 body 直传

- delegate wire format 从 JSON body 改为 HTTP headers 传递元数据
  (X-Delegate-Method/Url/Headers)，上游请求体 gzip 压缩后直传，
  减少跨国代理传输耗时
- Rust 侧新增 gzip 解压（含 decompression bomb 防护 50MB 上限）
- 升级模块从 GitHub API asset 下载改为公开 release URL 直链，
  GITHUB_TOKEN 变为可选
- 前端适配新 timing 字段，展示压缩率与 wire_size/body_size
- CI release notes 改用 generate_release_notes 自动生成

.github/workflows/build-proxy.yml

@@ -108,42 +108,7 @@ jobs:
         uses: softprops/action-gh-release@v2
         with:
           name: "aether-proxy ${{ github.ref_name }}"
-          body: |
+          generate_release_notes: true
-            ## aether-proxy ${{ github.ref_name }}
-
-            ### 下载
-
-            | 平台 | 文件 |
-            |------|------|
-            | Linux x86_64 | `aether-proxy-linux-amd64.tar.gz` |
-            | Linux ARM64 | `aether-proxy-linux-arm64.tar.gz` |
-            | macOS x86_64 (Intel) | `aether-proxy-macos-amd64.tar.gz` |
-            | macOS ARM64 (Apple Silicon) | `aether-proxy-macos-arm64.tar.gz` |
-            | Windows x86_64 | `aether-proxy-windows-amd64.zip` |
-
-            ### 使用
-
-            ```bash
-            # Linux/macOS: 解压
-            tar xzf aether-proxy-<platform>.tar.gz
-
-            # Windows: 解压 zip 文件
-
-            # 配置环境变量 (或写入 .env 文件)
-            export AETHER_PROXY_AETHER_URL=https://your-aether.example.com
-            export AETHER_PROXY_MANAGEMENT_TOKEN=ae_xxx
-            export AETHER_PROXY_HMAC_KEY=your-hmac-key
-
-            # 运行
-            ./aether-proxy
-            ```
-
-            ### 校验
-
-            下载 `SHA256SUMS.txt` 后可验证文件完整性:
-            ```bash
-            sha256sum -c SHA256SUMS.txt
-            ```
           files: |
             artifacts/aether-proxy-*
             artifacts/SHA256SUMS.txt

aether-proxy/Cargo.lock

@@ -10,7 +10,7 @@ checksum = "320119579fcad9c21884f5c4861d16174d0e06250625266f50fe6898340abefa"
 
 [[package]]
 name = "aether-proxy"
-version = "0.1.0"
+version = "0.1.2"
 dependencies = [
  "anyhow",
  "base64",

aether-proxy/src/proxy/delegate.rs

@@ -1,5 +1,6 @@
 use std::collections::HashMap;
 use std::collections::HashSet;
+use std::io::Read;
 use std::sync::Arc;
 use std::time::Instant;
 
@@ -7,7 +8,6 @@ use futures_util::TryStreamExt;
 use http_body_util::{BodyExt, Full, Limited, StreamBody};
 use hyper::body::{Frame, Incoming};
 use hyper::{Request, Response};
-use serde::Deserialize;
 use tracing::{debug, warn};
 use url::Url;
 
@@ -16,22 +16,20 @@ use crate::auth;
 use crate::config::Config;
 use crate::proxy::target_filter::{self, DnsCache};
 
-/// Delegation request payload sent by Aether.
-#[derive(Debug, Deserialize)]
-struct DelegateRequest {
-    method: String,
-    url: String,
-    headers: HashMap<String, String>,
-    body: Option<String>,
-    /// Accepted but not used on the proxy side — Aether controls timeouts.
-    #[allow(dead_code)]
-    timeout: Option<u64>,
-}
-
 /// Handle delegation requests: Aether sends a full request description,
 /// and the proxy issues the actual upstream HTTP call using its own TLS stack.
 ///
 /// Endpoint: POST /_aether/delegate
+///
+/// Wire format: metadata in HTTP headers, upstream body sent directly
+/// as HTTP body (optionally gzip-compressed via `Content-Encoding: gzip`).
+///
+/// Headers:
+///   X-Delegate-Method: POST
+///   X-Delegate-Url: https://api.anthropic.com/v1/messages
+///   X-Delegate-Headers: base64-encoded JSON {"Authorization": "Bearer ...", ...}
+///   X-Delegate-Timeout: 30              (accepted but not used — Aether controls timeouts)
+///   Content-Encoding: gzip              (optional, indicates body is gzip-compressed)
 pub async fn handle_delegate(
     req: Request<Incoming>,
     config: Arc<Config>,
@@ -42,7 +40,7 @@ pub async fn handle_delegate(
 ) -> Response<BoxBody> {
     let total_start = Instant::now();
 
-    // Authenticate via Authorization header (same HMAC scheme as Proxy-Authorization)
+    // ── Auth ──
     let auth_header = req
         .headers()
         .get("authorization")
@@ -54,34 +52,74 @@ pub async fn handle_delegate(
     }
     let auth_ms = total_start.elapsed().as_millis() as u64;
 
-    // Read and parse request body (limit to 10 MB to prevent OOM)
+    // ── Parse metadata from headers ──
-    const MAX_BODY: usize = 10 * 1024 * 1024;
+    let meta_start = Instant::now();
-    let body_read_start = Instant::now();
+
-    let body_bytes = match Limited::new(req.into_body(), MAX_BODY).collect().await {
+    let method_str = match req
-        Ok(collected) => collected.to_bytes(),
+        .headers()
-        Err(e) => {
+        .get("x-delegate-method")
-            warn!(error = %e, "failed to read delegate request body");
+        .and_then(|v| v.to_str().ok())
-            return error_response(413, "payload_too_large", "request body exceeds 10MB limit");
+    {
+        Some(m) => m.to_string(),
+        None => {
+            warn!("delegate missing X-Delegate-Method");
+            return error_response(400, "bad_request", "missing X-Delegate-Method header");
         }
     };
-    let body_read_ms = body_read_start.elapsed().as_millis() as u64;
-    let body_size = body_bytes.len() as u64;
 
-    let body_parse_start = Instant::now();
+    let target_url = match req
-    let delegate_req: DelegateRequest = match serde_json::from_slice(&body_bytes) {
+        .headers()
-        Ok(r) => r,
+        .get("x-delegate-url")
-        Err(e) => {
+        .and_then(|v| v.to_str().ok())
-            warn!(error = %e, "invalid delegate request JSON");
+    {
-            return error_response(400, "bad_request", &format!("invalid JSON: {}", e));
+        Some(u) => u.to_string(),
+        None => {
+            warn!("delegate missing X-Delegate-Url");
+            return error_response(400, "bad_request", "missing X-Delegate-Url header");
         }
     };
-    let body_parse_ms = body_parse_start.elapsed().as_millis() as u64;
 
-    // Target filter: validate the upstream URL against allowed ports and private IP checks
+    let upstream_headers: HashMap<String, String> = match req
-    let parsed_url = match Url::parse(&delegate_req.url) {
+        .headers()
+        .get("x-delegate-headers")
+        .and_then(|v| v.to_str().ok())
+    {
+        Some(b64) => {
+            match base64::Engine::decode(&base64::engine::general_purpose::STANDARD, b64) {
+                Ok(decoded) => match serde_json::from_slice(&decoded) {
+                    Ok(h) => h,
+                    Err(e) => {
+                        warn!(error = %e, "delegate invalid X-Delegate-Headers JSON");
+                        return error_response(
+                            400,
+                            "bad_request",
+                            "invalid X-Delegate-Headers JSON",
+                        );
+                    }
+                },
+                Err(e) => {
+                    warn!(error = %e, "delegate invalid X-Delegate-Headers base64");
+                    return error_response(400, "bad_request", "invalid X-Delegate-Headers base64");
+                }
+            }
+        }
+        None => HashMap::new(),
+    };
+
+    let is_gzip = req
+        .headers()
+        .get("content-encoding")
+        .and_then(|v| v.to_str().ok())
+        .map(|v| v.eq_ignore_ascii_case("gzip"))
+        .unwrap_or(false);
+
+    let meta_ms = meta_start.elapsed().as_millis() as u64;
+
+    // ── Target validation ──
+    let parsed_url = match Url::parse(&target_url) {
         Ok(u) => u,
         Err(e) => {
-            warn!(url = %delegate_req.url, error = %e, "invalid delegate target URL");
+            warn!(url = %target_url, error = %e, "delegate invalid target URL");
             return error_response(400, "bad_request", &format!("invalid URL: {}", e));
         }
     };
@@ -89,7 +127,7 @@ pub async fn handle_delegate(
     let host = match parsed_url.host_str() {
         Some(h) => h.to_string(),
         None => {
-            warn!(url = %delegate_req.url, "delegate target URL missing host");
+            warn!(url = %target_url, "delegate target URL missing host");
             return error_response(400, "bad_request", "URL missing host");
         }
     };
@@ -103,51 +141,91 @@ pub async fn handle_delegate(
     }
     let dns_ms = dns_start.elapsed().as_millis() as u64;
 
-    debug!(
+    debug!(method = %method_str, url = %target_url, is_gzip, "delegate request");
-        method = %delegate_req.method,
-        url = %delegate_req.url,
-        "delegate request"
-    );
 
-    // Build upstream request
+    // ── Read body + decompress if gzip ──
-    let method = match delegate_req.method.parse::<reqwest::Method>() {
+    let body_read_start = Instant::now();
+    const MAX_BODY: usize = 10 * 1024 * 1024;
+    let body_bytes = match Limited::new(req.into_body(), MAX_BODY).collect().await {
+        Ok(collected) => collected.to_bytes(),
+        Err(e) => {
+            warn!(error = %e, "delegate failed to read request body");
+            return error_response(413, "payload_too_large", "request body exceeds 10MB limit");
+        }
+    };
+    let body_read_ms = body_read_start.elapsed().as_millis() as u64;
+    let wire_size = body_bytes.len() as u64;
+
+    let decompress_start = Instant::now();
+    const MAX_DECOMPRESSED: usize = 50 * 1024 * 1024; // 50 MB
+    let upstream_body: Option<Vec<u8>> = if body_bytes.is_empty() {
+        None
+    } else if is_gzip {
+        let decoder = flate2::read::GzDecoder::new(&body_bytes[..]);
+        let mut decompressed = Vec::with_capacity((body_bytes.len() * 4).min(MAX_DECOMPRESSED));
+        match decoder
+            .take(MAX_DECOMPRESSED as u64 + 1)
+            .read_to_end(&mut decompressed)
+        {
+            Ok(n) if n > MAX_DECOMPRESSED => {
+                warn!(
+                    wire_size = body_bytes.len(),
+                    decompressed_size = n,
+                    "delegate decompressed body exceeds limit"
+                );
+                return error_response(
+                    413,
+                    "payload_too_large",
+                    "decompressed body exceeds 50MB limit",
+                );
+            }
+            Ok(_) => Some(decompressed),
+            Err(e) => {
+                warn!(error = %e, "delegate gzip decompression failed");
+                return error_response(400, "bad_request", "gzip decompression failed");
+            }
+        }
+    } else {
+        Some(body_bytes.to_vec())
+    };
+    let decompress_ms = decompress_start.elapsed().as_millis() as u64;
+    let body_size = upstream_body.as_ref().map(|b| b.len() as u64).unwrap_or(0);
+
+    // ── Build upstream request ──
+    let method = match method_str.parse::<reqwest::Method>() {
         Ok(m) => m,
         Err(e) => {
-            warn!(error = %e, method = %delegate_req.method, "invalid HTTP method");
+            warn!(error = %e, method = %method_str, "delegate invalid HTTP method");
             return error_response(400, "bad_request", &format!("invalid method: {}", e));
         }
     };
 
-    let mut upstream_req = http_client.request(method, &delegate_req.url);
+    let mut upstream_req = http_client.request(method, &target_url);
-
-    // NOTE: We intentionally do NOT set a per-request timeout here.
-    // reqwest's `.timeout()` caps the *entire* request including body streaming,
-    // which would truncate long-lived SSE streams. The delegate_client already
-    // has a configured connect_timeout for connection establishment, and Aether controls
-    // first-byte / idle timeouts on its own side via asyncio.
 
     // Set headers (skip `host` — reqwest sets it from the URL automatically,
     // and a duplicate Host header can confuse certain upstreams)
-    for (name, value) in &delegate_req.headers {
+    for (name, value) in &upstream_headers {
         if name.eq_ignore_ascii_case("host") {
             continue;
         }
         upstream_req = upstream_req.header(name.as_str(), value.as_str());
     }
 
-    // Set body
+    if let Some(body) = upstream_body {
-    if let Some(body) = delegate_req.body {
         upstream_req = upstream_req.body(body);
     }
 
-    // Send upstream request
+    // ── Send upstream request ──
+    // NOTE: We intentionally do NOT set a per-request timeout here.
+    // reqwest's `.timeout()` caps the *entire* request including body streaming,
+    // which would truncate long-lived SSE streams. The delegate_client already
+    // has a configured connect_timeout for connection establishment, and Aether controls
+    // first-byte / idle timeouts on its own side via asyncio.
     let upstream_start = Instant::now();
     let upstream_resp = match upstream_req.send().await {
         Ok(resp) => resp,
         Err(e) => {
-            warn!(url = %delegate_req.url, error = %e, "delegate upstream request failed");
+            warn!(url = %target_url, error = %e, "delegate upstream request failed");
-            // Sanitize: strip URL details from error message to avoid leaking
-            // API keys or paths that may appear in query strings / paths.
             let safe_detail = sanitize_upstream_error(&e.to_string());
             if e.is_timeout() {
                 return error_response(504, "upstream_timeout", &safe_detail);
@@ -157,33 +235,34 @@ pub async fn handle_delegate(
     };
     let upstream_ms = upstream_start.elapsed().as_millis() as u64;
 
-    // Build response: pass through upstream status + headers, stream body back
+    // ── Build response ──
     let status = upstream_resp.status().as_u16();
-    let upstream_headers = upstream_resp.headers().clone();
+    let resp_headers = upstream_resp.headers().clone();
-
     let total_ms = total_start.elapsed().as_millis() as u64;
 
     debug!(
-        url = %delegate_req.url,
+        url = %target_url,
         status,
         dns_ms,
         upstream_ms,
         total_ms,
+        wire_size,
+        body_size,
         "delegate upstream response"
     );
 
-    // Inject proxy timing header for Aether to parse
     let timing = serde_json::json!({
         "auth_ms": auth_ms,
+        "meta_ms": meta_ms,
         "body_read_ms": body_read_ms,
-        "body_parse_ms": body_parse_ms,
+        "decompress_ms": decompress_ms,
+        "wire_size": wire_size,
         "body_size": body_size,
         "dns_ms": dns_ms,
         "upstream_ms": upstream_ms,
         "total_ms": total_ms,
     });
 
-    // Stream the response body
     let body_stream = upstream_resp
         .bytes_stream()
         .map_ok(Frame::data)
@@ -192,7 +271,7 @@ pub async fn handle_delegate(
     let stream_body: BoxBody = StreamBody::new(body_stream).boxed();
 
     let mut builder = Response::builder().status(status);
-    for (name, value) in upstream_headers.iter() {
+    for (name, value) in resp_headers.iter() {
         builder = builder.header(name, value);
     }
     builder = builder.header("X-Proxy-Timing", timing.to_string());

aether-proxy/src/setup/upgrade.rs

@@ -17,13 +17,6 @@ const CURRENT_VERSION: &str = env!("CARGO_PKG_VERSION");
 struct GithubRelease {
     tag_name: String,
     name: String,
-    assets: Vec<GithubAsset>,
-}
-
-#[derive(serde::Deserialize)]
-struct GithubAsset {
-    name: String,
-    id: u64,
 }
 
 // ── Platform detection ───────────────────────────────────────────────────────
@@ -49,17 +42,15 @@ fn detect_platform() -> &'static str {
 // ── GitHub HTTP client ───────────────────────────────────────────────────────
 
 fn build_github_client() -> anyhow::Result<reqwest::Client> {
-    let token = std::env::var("GITHUB_TOKEN").map_err(|_| {
-        anyhow::anyhow!(
-            "GITHUB_TOKEN is required (private repo).\n  Set it via: export GITHUB_TOKEN=ghp_xxx"
-        )
-    })?;
-
     let mut headers = reqwest::header::HeaderMap::new();
-    headers.insert(
+
-        reqwest::header::AUTHORIZATION,
+    if let Ok(token) = std::env::var("GITHUB_TOKEN") {
-        reqwest::header::HeaderValue::from_str(&format!("Bearer {}", token))?,
+        headers.insert(
-    );
+            reqwest::header::AUTHORIZATION,
+            reqwest::header::HeaderValue::from_str(&format!("Bearer {}", token))?,
+        );
+    }
+
     headers.insert(
         reqwest::header::ACCEPT,
         reqwest::header::HeaderValue::from_static("application/vnd.github+json"),
@@ -119,16 +110,18 @@ async fn fetch_release(
     }
 }
 
-// ── Download & verify ────────────────────────────────────────────────────────
+// ── Download via GitHub release direct links ─────────────────────────────────
 
-async fn download_asset_bytes(
+/// Download a release asset via the public direct download URL:
+/// `https://github.com/{repo}/releases/download/{tag}/{filename}`
+async fn download_release_file(
     client: &reqwest::Client,
-    asset: &GithubAsset,
+    tag: &str,
+    filename: &str,
 ) -> anyhow::Result<Vec<u8>> {
-    // Use GitHub API asset endpoint for reliable private repo downloads
     let url = format!(
-        "{}/repos/{}/releases/assets/{}",
+        "https://github.com/{}/releases/download/{}/{}",
-        GITHUB_API_BASE, GITHUB_REPO, asset.id
+        GITHUB_REPO, tag, filename
     );
     let resp = client
         .get(&url)
@@ -138,7 +131,7 @@ async fn download_asset_bytes(
     if !resp.status().is_success() {
         anyhow::bail!(
             "download failed for '{}' (HTTP {})",
-            asset.name,
+            filename,
             resp.status(),
         );
     }
@@ -161,28 +154,16 @@ fn parse_checksum(sums_text: &str, filename: &str) -> anyhow::Result<String> {
 
 async fn download_and_verify(
     client: &reqwest::Client,
-    release: &GithubRelease,
+    tag: &str,
     platform: &str,
     dest: &Path,
 ) -> anyhow::Result<()> {
     let archive_name = format!("aether-proxy-{}.tar.gz", platform);
 
-    let archive_asset = release
-        .assets
-        .iter()
-        .find(|a| a.name == archive_name)
-        .ok_or_else(|| anyhow::anyhow!("asset '{}' not found in release", archive_name))?;
-
-    let checksum_asset = release
-        .assets
-        .iter()
-        .find(|a| a.name == "SHA256SUMS.txt")
-        .ok_or_else(|| anyhow::anyhow!("SHA256SUMS.txt not found in release"))?;
-
     eprintln!("  Downloading {}...", archive_name);
     let (archive_bytes, checksum_bytes) = tokio::try_join!(
-        download_asset_bytes(client, archive_asset),
+        download_release_file(client, tag, &archive_name),
-        download_asset_bytes(client, checksum_asset),
+        download_release_file(client, tag, "SHA256SUMS.txt"),
     )?;
     let checksum_text = String::from_utf8(checksum_bytes)?;
 
@@ -348,7 +329,7 @@ pub async fn cmd_upgrade(version: Option<String>) -> anyhow::Result<()> {
     eprintln!("  Upgrading: {} -> {}", CURRENT_VERSION, target_semver);
     eprintln!();
 
-    if let Err(e) = download_and_verify(&client, &release, platform, &temp_path).await {
+    if let Err(e) = download_and_verify(&client, target_tag, platform, &temp_path).await {
         let _ = std::fs::remove_file(&temp_path);
         return Err(e);
     }

frontend/src/features/usage/components/HorizontalRequestTimeline.vue

@@ -452,6 +452,13 @@ const formatLatency = (ms: number | undefined | null): string => {
   return `${ms}ms`
 }
 
+// 格式化字节大小
+const formatSize = (bytes: number): string => {
+  if (bytes >= 1048576) return `${(bytes / 1048576).toFixed(1)}MB`
+  if (bytes >= 1024) return `${(bytes / 1024).toFixed(1)}KB`
+  return `${bytes}B`
+}
+
 // 代理 timing 分阶段展示
 const proxyTimingBreakdown = (proxy: Record<string, any>): string => {
   const t = proxy.timing
@@ -459,9 +466,16 @@ const proxyTimingBreakdown = (proxy: Record<string, any>): string => {
 
   const parts: string[] = []
 
-  // 新版细分字段（aether-proxy 新版本上报）
+  // 读取 + 解压合并显示（含压缩率）
-  if (t.body_read_ms != null && t.body_read_ms > 0) {
+  const readDecompress = (t.body_read_ms || 0) + (t.decompress_ms || 0)
-    parts.push(`读取 ${formatLatency(t.body_read_ms)}`)
+  if (readDecompress > 0) {
+    let label = `读取 ${formatLatency(readDecompress)}`
+    if (t.decompress_ms != null && t.decompress_ms > 0 && t.wire_size != null && t.body_size != null && t.body_size > 0) {
+      const ratio = Math.round((1 - t.wire_size / t.body_size) * 100)
+      label += ` ${formatSize(t.wire_size)}→${formatSize(t.body_size)}`
+      if (ratio > 0) label += ` -${ratio}%`
+    }
+    parts.push(label)
   }
 
   parts.push(`DNS ${formatLatency(t.dns_ms)}`)
@@ -471,7 +485,6 @@ const proxyTimingBreakdown = (proxy: Record<string, any>): string => {
   if (proxy.ttfb_ms != null && t.total_ms != null) {
     const gap = proxy.ttfb_ms - t.total_ms
     if (gap > 500) {
-      // 超过 500ms 的差值才显示，排除正常网络 RTT
       parts.push(`传输 ${formatLatency(Math.round(gap))}`)
     }
   }

src/services/proxy_node/resolver.py

@@ -8,8 +8,10 @@
 from __future__ import annotations
 
 import base64
+import gzip as _gzip
 import hashlib
 import hmac as _hmac
+import json as _json
 import time
 from typing import Any
 from urllib.parse import quote, urlparse
@@ -613,6 +615,9 @@ def _build_delegate_kwargs_core(
     """
     构建代发请求的核心参数（post/stream 共用）
 
+    元数据通过 HTTP headers 传递（X-Delegate-Method/Url/Headers），
+    上游请求体 gzip 压缩后直接作为 HTTP body 发送，大幅减少跨国传输耗时。
+
     Args:
         delegate_cfg: resolve_delegate_config 返回的配置
         url:          上游实际 URL
@@ -621,27 +626,42 @@ def _build_delegate_kwargs_core(
         timeout:      上游超时秒数
         refresh_auth: 为 True 时重新生成 HMAC 签名（用于 retry）
     """
-    import json as _json
-
     auth = (
         delegate_cfg["fresh_auth_header"]()
         if refresh_auth
         else delegate_cfg.get("auth_header") or delegate_cfg["fresh_auth_header"]()
     )
 
-    return {
+    # 上游 headers base64 编码
+    headers_b64 = base64.b64encode(_json.dumps(headers, ensure_ascii=False).encode("utf-8")).decode(
+        "ascii"
+    )
+
+    # 构建代发请求 headers（元数据）
+    delegate_headers: dict[str, str] = {
+        "Authorization": auth,
+        "X-Delegate-Method": "POST",
+        "X-Delegate-Url": url,
+        "X-Delegate-Headers": headers_b64,
+        "X-Delegate-Timeout": str(int(timeout)),
+    }
+
+    kwargs: dict[str, Any] = {
         "url": delegate_cfg["delegate_url"],
-        "json": {
+        "headers": delegate_headers,
-            "method": "POST",
-            "url": url,
-            "headers": headers,
-            "body": _json.dumps(payload, ensure_ascii=False) if payload is not None else None,
-            "timeout": int(timeout),
-        },
-        "headers": {"Authorization": auth, "Content-Type": _JSON_CT},
         "timeout": httpx.Timeout(timeout + 10),
     }
 
+    # body gzip 压缩后直接作为 HTTP content
+    if payload is not None:
+        body_bytes = _json.dumps(payload, ensure_ascii=False).encode("utf-8")
+        compressed = _gzip.compress(body_bytes)
+        kwargs["content"] = compressed
+        kwargs["headers"]["Content-Encoding"] = "gzip"
+        kwargs["headers"]["Content-Type"] = _JSON_CT
+
+    return kwargs
+
 
 def build_delegate_post_kwargs(
     delegate_cfg: dict[str, Any],