feat: 启用响应压缩(gzip/brotli/zstd) + 优化SSE流式chunk_size

- 添加 flask-compress 中间件，长文本响应自动压缩（压缩率70-80%）
- 压缩阈值设为256字节，小响应不压缩
- SSE流式 chunk_size 从 4096 提升到 16384，减少分片开销
- 压缩白名单包含 text/event-stream，SSE也支持压缩
- 解决长文本响应体积过大导致超时/内存溢出的问题

ai/huawei_gateway.py

@@ -21,6 +21,13 @@ from concurrent.futures import ThreadPoolExecutor, as_completed
 
 from flask import Flask, request, Response
 
+# 尝试导入 flask-compress，用于响应压缩
+try:
+    from flask_compress import Compress
+    _compress_available = True
+except ImportError:
+    _compress_available = False
+
 # 尝试导入 requests，失败则给出明确提示
 try:
     import requests
@@ -241,6 +248,19 @@ http_session.mount('http://', adapter)
 # ================= Flask 应用 =================
 app = Flask(__name__)
 
+# 启用响应压缩（gzip/brotli/zstd），对长文本响应压缩率可达 70-80%
+if _compress_available:
+    compress = Compress()
+    compress.init_app(app)
+    # 设置压缩最小阈值（256字节以下不压缩）
+    app.config['COMPRESS_MIN_SIZE'] = 256
+    # 压缩 MIME 类型白名单
+    app.config['COMPRESS_MIMETYPES'] = [
+        'text/html', 'text/plain', 'text/css', 'text/xml',
+        'application/json', 'application/javascript',
+        'application/xml', 'text/event-stream',
+    ]
+
 @app.route('/health')
 def health():
     """健康检查端点"""
@@ -350,7 +370,7 @@ def proxy(subpath):
         if 'text/event-stream' in content_type or resp.headers.get('Transfer-Encoding', '') == 'chunked':
             def sse_stream():
                 try:
-                    for chunk in resp.iter_content(chunk_size=4096):
+                    for chunk in resp.iter_content(chunk_size=16384):
                         if chunk:
                             yield chunk
                 finally: