refactor: split streaming endpoint from Azure Functions

Disable /search/stream in Function runtime with a 501 guidance response, keep streaming on always-on FastAPI service, and update README with the new deployment and routing model.

Made-with: Cursor

README.md

@@ -103,9 +103,10 @@ uvicorn serve:app --host 0.0.0.0 --port 8080 --reload
 |------|------|------|
 | GET | `/health` | 健康检查（本地 FastAPI） |
 | POST | `/search` | 同步搜索（阻塞式，本地 FastAPI） |
-| POST | `/search/stream` | 流式搜索（SSE 事件，本地 FastAPI） |
+| POST | `/search/stream` | 流式搜索（SSE 事件，仅常驻 FastAPI 服务） |
 
-> Azure Functions 部署时由 `function_app.py` 挂载到 `/api` 前缀，对外路径分别为 `/api/health`、`/api/search`、`/api/search/stream`。
+> Azure Functions 部署时由 `function_app.py` 挂载到 `/api` 前缀，仅提供 `/api/health`、`/api/search`。  
+> `/search/stream` 必须走常驻 `uvicorn` 服务（避免 Functions 网关缓冲导致假流式）。
 
 ### 1. 健康检查
 
@@ -144,10 +145,10 @@ Jina 搜索结果固定上限：
 - `cache_hit` - 是否缓存命中
 - `audit_id` - 审计追踪 ID
 
-### 3. SSE 流式搜索
+### 3. SSE 流式搜索（常驻 FastAPI）
 
 ```bash
-curl -N -X POST https://aisousuo.azurewebsites.net/api/search/stream \
+curl -N -X POST https://your-stream-service/search/stream \
   -H "Content-Type: application/json" \
   -d '{"query":"PydanticAI 适合哪些场景?","search_mode":"deep"}'
 ```
@@ -157,6 +158,8 @@ curl -N -X POST https://aisousuo.azurewebsites.net/api/search/stream \
 
 > 当请求超时时，SSE 会发送 `timeout` 事件并结束流。
 
+> 如果你误调 Azure Functions 的 `/api/search/stream`，会返回 `501`，并提示使用独立流式服务地址。
+
 ---
 
 ## 环境变量配置
@@ -233,6 +236,17 @@ func azure functionapp publish aisousuo --python
 curl https://aisousuo.azurewebsites.net/api/health
 ```
 
+### 流式服务部署（推荐 App Service / Container）
+
+```bash
+# 启动常驻服务（示例）
+uvicorn serve:app --host 0.0.0.0 --port 8080
+```
+
+- 将前端流式请求改为 `https://<stream-service>/search/stream`
+- Azure Functions 继续承载同步接口：`/api/search`
+- 可在 Functions 环境变量中配置 `STREAM_SERVICE_BASE_URL=https://<stream-service>`，用于错误提示回传目标流式地址
+
 ---
 
 ## 设计架构

ai_search_agent/api.py

@@ -2,6 +2,7 @@ from __future__ import annotations
 
 import asyncio
 import json
+import os
 
 from fastapi import FastAPI
 from fastapi.responses import JSONResponse
@@ -13,7 +14,7 @@ from .config import Settings
 from .models import SearchRequest, SearchResponse
 
 
-def create_app() -> FastAPI:
+def create_app(enable_streaming: bool = True) -> FastAPI:
     settings = Settings.from_env()
     agent = AISearchAgent(settings)
     audit_logger = AuditLogger(settings.audit_log_path)
@@ -44,7 +45,22 @@ def create_app() -> FastAPI:
         return JSONResponse(content=response.model_dump())
 
     @app.post("/search/stream")
-    async def search_stream(request: SearchRequest) -> StreamingResponse:
+    async def search_stream(request: SearchRequest):
+        if not enable_streaming:
+            stream_base_url = os.getenv("STREAM_SERVICE_BASE_URL", "").rstrip("/")
+            return JSONResponse(
+                status_code=501,
+                content={
+                    "error": "STREAMING_NOT_AVAILABLE_IN_FUNCTIONS",
+                    "message": "Use the always-on stream service for /search/stream.",
+                    "stream_endpoint": (
+                        f"{stream_base_url}/search/stream" if stream_base_url else None
+                    ),
+                    "query": request.query,
+                    "search_mode": agent.resolve_search_mode(request),
+                },
+            )
+
         async def event_generator():
             # Send an immediate event so clients/proxies flush the SSE channel early.
             yield "event: connected\ndata: {}\n\n"

function_app.py

@@ -6,7 +6,9 @@ from starlette.routing import Mount
 
 from ai_search_agent.api import create_app
 
-_fastapi = create_app()
+# Azure Functions keeps only synchronous search endpoints.
+# Stream endpoint should be served by always-on uvicorn service.
+_fastapi = create_app(enable_streaming=False)
 
 # Azure Functions delivers the full path including /api/ prefix to the ASGI app.
 # Mount FastAPI under /api so that /api/health -> FastAPI /health, etc.