拆分前后端，后端用Python重写

README.md

@@ -1,7 +1,7 @@
 # 交互式AIGC场景 AIGC Demo
 
 此 Demo 为简化版本, 如您有 1.5.x 版本 UI 的诉求, 可切换至 1.5.1 分支。
-跑通阶段时, 无须关心代码实现，仅需按需完成 `Server/scenes/*.json` 的场景信息填充即可。
+跑通阶段时, 无须关心代码实现，仅需按需完成 `backend/scenes/*.json` 的场景信息填充即可。
 
 ## 简介
 - 在 AIGC 对话场景下，火山引擎 AIGC-RTC Server 云端服务，通过整合 RTC 音视频流处理，ASR 语音识别，大模型接口调用集成，以及 TTS 语音生成等能力，提供基于流式语音的端到端AIGC能力链路。
@@ -9,16 +9,20 @@
 - 同时火山引擎 RTC拥有成熟的音频 3A 处理、视频处理等技术以及大规模音视频聊天能力，可支持 AIGC 产品更便捷的支持多模态交互、多人互动等场景能力，保持交互的自然性和高效性。 
 
 ## 【必看】环境准备
-**Node 版本: 16.0+**
+
+> 本项目已重构为 monorepo 结构，前端位于 `frontend/`，Python 后端位于 `backend/`。
+
+**前端环境：Node 16.0+**
+**后端环境：Python 3.9+**
 
 ### 1. 运行环境
-需要准备两个 Terminal，分别启动服务端和前端页面。
+需要准备两个 Terminal，分别启动后端服务和前端页面。
 
 ### 2. 服务开通
 开通 ASR、TTS、LLM、RTC 等服务，可参考 [开通服务](https://www.volcengine.com/docs/6348/1315561?s=g) 进行相关服务的授权与开通。
 
 ### 3. 场景配置
-`Server/scenes/*.json`
+`backend/scenes/*.json`
 
 您可以自定义具体场景, 并按需根据模版填充 `SceneConfig`、`AccountConfig`、`RTCConfig`、`VoiceChat` 中需要的参数。
 
@@ -35,27 +39,18 @@ Demo 中以 `Custom` 场景为例，您可以自行新增场景。
     - 可通过 [快速跑通 Demo](https://console.volcengine.com/rtc/aigc/run?s=g) 快速获取参数, 跑通后点击右上角 `接入 API` 按钮复制相关代码贴到 JSON 配置文件中即可。
 ## 快速开始
 请注意，服务端和 Web 端都需要启动, 启动步骤如下:
-### 服务端
+### 后端服务（Python FastAPI）
-进到项目根目录
-#### 安装依赖
 ```shell
-cd Server
+cd backend
-yarn
+pip install -r requirements.txt
-```
+uvicorn main:app --host 0.0.0.0 --port 3001 --reload
-#### 运行项目
-```shell
-yarn dev
 ```
 
 ### 前端页面
-进到项目根目录
-#### 安装依赖
 ```shell
-yarn
+cd frontend
-```
+npm install
-#### 运行项目
+npm run dev
-```shell
-yarn dev
 ```
 
 ### 常见问题
@@ -67,7 +62,7 @@ yarn dev
 | **[StartVoiceChat]Failed(Reason: The task has been started. Please do not call the startup task interface repeatedly.)** 报错 | 如果设置的 RoomId、UserId 为固定值，重复调用 startAgent 会导致出错，只需先调用 stopAgent 后再重新 startAgent 即可。 |
 | 为什么麦克风、摄像头开启失败？浏览器报了`TypeError: Cannot read properties of undefined (reading 'getUserMedia')` | 检查当前页面是否为[安全上下文](https://developer.mozilla.org/zh-CN/docs/Web/Security/Secure_Contexts)（简单来说，检查当前页面是否为 `localhost` 或者 是否为 https 协议）。浏览器[限制](https://developer.mozilla.org/zh-CN/docs/Web/Security/Secure_Contexts/features_restricted_to_secure_contexts) `getUserMedia` 只能在安全上下文中使用。 |
 | 为什么我的麦克风正常、摄像头也正常，但是设备没有正常工作? | 可能是设备权限未授予，详情可参考 [Web 排查设备权限获取失败问题](https://www.volcengine.com/docs/6348/1356355?s=g)。 |
-| 接口调用时, 返回 "Invalid 'Authorization' header, Pls check your authorization header" 错误 | `Server/app.js` 中的 AK/SK 不正确 |
+| 接口调用时, 返回 "Invalid 'Authorization' header, Pls check your authorization header" 错误 | `backend/scenes/*.json` 中的 AK/SK 不正确 |
 | 什么是 RTC | **R**eal **T**ime **C**ommunication, RTC 的概念可参考[官网文档](https://www.volcengine.com/docs/6348/66812?s=g)。 |
 | 不清楚什么是主账号，什么是子账号 | 可以参考[官方概念](https://www.volcengine.com/docs/6257/64963?hyperlink_open_type=lark.open_in_browser&s=g) 。|
 | 我有自己的服务端了, 我应该怎么让前端调用我的服务端呢 | 修改 `src/config/index.ts` 中的 `AIGC_PROXY_HOST` 请求域名和接口并在 `src/app/api.ts` 中修改接口参数配置 `APIS_CONFIG` |

backend/README.md

@@ -0,0 +1,49 @@
+# AIGC Backend (Python FastAPI)
+
+原 Node.js + Koa 服务的 Python 重写版本，使用 FastAPI 框架。
+
+## 环境要求
+
+- Python 3.9+
+
+## 安装依赖
+
+```shell
+pip install -r requirements.txt
+```
+
+## 场景配置
+
+编辑 `scenes/*.json`，填写以下字段：
+
+| 字段 | 说明 |
+|------|------|
+| `AccountConfig.accessKeyId` | 火山引擎 AK，从 https://console.volcengine.com/iam/keymanage/ 获取 |
+| `AccountConfig.secretKey` | 火山引擎 SK |
+| `RTCConfig.AppId` | RTC 应用 ID |
+| `RTCConfig.AppKey` | RTC 应用 Key（用于自动生成 Token） |
+| `VoiceChat.*` | AIGC 相关配置，参考 https://www.volcengine.com/docs/6348/1558163 |
+
+## 启动服务
+
+```shell
+uvicorn main:app --host 0.0.0.0 --port 3001 --reload
+```
+
+服务启动后监听 `http://localhost:3001`。
+
+## 接口说明
+
+### POST /getScenes
+
+返回所有场景列表，自动生成 RoomId/UserId/Token（若未在 JSON 中配置）。
+
+### POST /proxy?Action={Action}&Version={Version}
+
+代理转发至火山引擎 RTC OpenAPI。
+
+支持的 Action：
+- `StartVoiceChat` — 启动语音对话
+- `StopVoiceChat` — 停止语音对话
+
+请求体需包含 `SceneID` 字段，对应 `scenes/` 目录下的 JSON 文件名（不含扩展名）。

backend/main.py

@@ -0,0 +1,190 @@
+"""
+Copyright 2025 Beijing Volcano Engine Technology Co., Ltd. All Rights Reserved.
+SPDX-license-identifier: BSD-3-Clause
+
+FastAPI backend — migrated from Server/app.js (Node.js + Koa)
+"""
+import json
+import os
+import time
+import uuid
+from pathlib import Path
+
+import httpx
+from fastapi import FastAPI, Request
+from fastapi.middleware.cors import CORSMiddleware
+from fastapi.responses import JSONResponse
+
+from signer import Signer
+from token import AccessToken, privileges
+
+app = FastAPI()
+
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+SCENES_DIR = Path(__file__).parent / "scenes"
+
+
+def load_scenes() -> dict:
+    scenes = {}
+    for p in SCENES_DIR.glob("*.json"):
+        with open(p, encoding="utf-8") as f:
+            scenes[p.stem] = json.load(f)
+    return scenes
+
+
+Scenes = load_scenes()
+
+
+def assert_value(value, msg: str):
+    if not value or (isinstance(value, str) and " " in value):
+        raise ValueError(msg)
+
+
+def error_response(action: str, message: str):
+    return JSONResponse({
+        "ResponseMetadata": {
+            "Action": action,
+            "Error": {"Code": -1, "Message": message},
+        }
+    })
+
+
+@app.post("/proxy")
+async def proxy(request: Request):
+    action = request.query_params.get("Action", "")
+    version = request.query_params.get("Version", "2024-12-01")
+
+    try:
+        assert_value(action, "Action 不能为空")
+        assert_value(version, "Version 不能为空")
+
+        body = await request.json()
+        scene_id = body.get("SceneID", "")
+        assert_value(scene_id, "SceneID 不能为空，SceneID 用于指定场景的 JSON")
+
+        json_data = Scenes.get(scene_id)
+        if not json_data:
+            raise ValueError(f"{scene_id} 不存在，请先在 backend/scenes 下定义该场景的 JSON.")
+
+        voice_chat = json_data.get("VoiceChat", {})
+        account_config = json_data.get("AccountConfig", {})
+        assert_value(account_config.get("accessKeyId"), "AccountConfig.accessKeyId 不能为空")
+        assert_value(account_config.get("secretKey"), "AccountConfig.secretKey 不能为空")
+
+        if action == "StartVoiceChat":
+            req_body = voice_chat
+        elif action == "StopVoiceChat":
+            app_id = voice_chat.get("AppId", "")
+            room_id = voice_chat.get("RoomId", "")
+            task_id = voice_chat.get("TaskId", "")
+            assert_value(app_id, "VoiceChat.AppId 不能为空")
+            assert_value(room_id, "VoiceChat.RoomId 不能为空")
+            assert_value(task_id, "VoiceChat.TaskId 不能为空")
+            req_body = {"AppId": app_id, "RoomId": room_id, "TaskId": task_id}
+        else:
+            req_body = {}
+
+        request_data = {
+            "region": "cn-north-1",
+            "method": "POST",
+            "params": {"Action": action, "Version": version},
+            "headers": {
+                "Host": "rtc.volcengineapi.com",
+                "Content-type": "application/json",
+            },
+            "body": req_body,
+        }
+        signer = Signer(request_data, "rtc")
+        signer.add_authorization(account_config)
+
+        async with httpx.AsyncClient() as client:
+            resp = await client.post(
+                f"https://rtc.volcengineapi.com?Action={action}&Version={version}",
+                headers=request_data["headers"],
+                json=req_body,
+            )
+        return JSONResponse(resp.json())
+
+    except ValueError as e:
+        return error_response(action, str(e))
+    except Exception as e:
+        return error_response(action, str(e))
+
+
+@app.post("/getScenes")
+async def get_scenes():
+    try:
+        scenes_list = []
+        for scene_name, data in Scenes.items():
+            scene_config = data.get("SceneConfig", {})
+            rtc_config = data.get("RTCConfig", {})
+            voice_chat = data.get("VoiceChat", {})
+
+            app_id = rtc_config.get("AppId", "")
+            assert_value(app_id, f"{scene_name} 场景的 RTCConfig.AppId 不能为空")
+
+            token = rtc_config.get("Token", "")
+            user_id = rtc_config.get("UserId", "")
+            room_id = rtc_config.get("RoomId", "")
+            app_key = rtc_config.get("AppKey", "")
+
+            if app_id and (not token or not user_id or not room_id):
+                rtc_config["RoomId"] = voice_chat["RoomId"] = room_id or str(uuid.uuid4())
+                rtc_config["UserId"] = user_id = user_id or str(uuid.uuid4())
+                if voice_chat.get("AgentConfig") and voice_chat["AgentConfig"].get("TargetUserId"):
+                    voice_chat["AgentConfig"]["TargetUserId"][0] = rtc_config["UserId"]
+
+                assert_value(app_key, f"自动生成 Token 时，{scene_name} 场景的 AppKey 不可为空")
+                key = AccessToken(app_id, app_key, rtc_config["RoomId"], rtc_config["UserId"])
+                key.add_privilege(privileges["PrivSubscribeStream"], 0)
+                key.add_privilege(privileges["PrivPublishStream"], 0)
+                key.expire_time(int(time.time()) + 24 * 3600)
+                rtc_config["Token"] = key.serialize()
+
+            scene_config["id"] = scene_name
+            scene_config["botName"] = voice_chat.get("AgentConfig", {}).get("UserId")
+            scene_config["isInterruptMode"] = voice_chat.get("Config", {}).get("InterruptMode") == 0
+            scene_config["isVision"] = (
+                voice_chat.get("Config", {}).get("LLMConfig", {}).get("VisionConfig", {}).get("Enable")
+            )
+            scene_config["isScreenMode"] = (
+                voice_chat.get("Config", {}).get("LLMConfig", {})
+                .get("VisionConfig", {}).get("SnapshotConfig", {}).get("StreamType") == 1
+            )
+            scene_config["isAvatarScene"] = (
+                voice_chat.get("Config", {}).get("AvatarConfig", {}).get("Enabled")
+            )
+            scene_config["avatarBgUrl"] = (
+                voice_chat.get("Config", {}).get("AvatarConfig", {}).get("BackgroundUrl")
+            )
+
+            rtc_out = {k: v for k, v in rtc_config.items() if k != "AppKey"}
+
+            scenes_list.append({
+                "scene": scene_config,
+                "rtc": rtc_out,
+            })
+
+        return JSONResponse({
+            "ResponseMetadata": {"Action": "getScenes"},
+            "Result": {"scenes": scenes_list},
+        })
+
+    except ValueError as e:
+        return JSONResponse({
+            "ResponseMetadata": {
+                "Action": "getScenes",
+                "Error": {"Code": -1, "Message": str(e)},
+            }
+        })
+
+
+if __name__ == "__main__":
+    import uvicorn
+    uvicorn.run("main:app", host="0.0.0.0", port=3001, reload=True)

backend/requirements.txt

@@ -0,0 +1,4 @@
+fastapi>=0.110.0
+uvicorn[standard]>=0.29.0
+httpx>=0.27.0
+python-multipart>=0.0.9

backend/scenes/Custom.json

@@ -0,0 +1,75 @@
+{
+  "SceneConfig": {
+    "icon": "https://lf3-rtc-demo.volccdn.com/obj/rtc-aigc-assets/DoubaoAvatar.png",
+    "name": "自定义助手"
+  },
+  "AccountConfig": {
+    "accessKeyId": "",
+    "secretKey": ""
+  },
+  "RTCConfig": {
+    "AppId": "",
+    "AppKey": "",
+    "RoomId": "",
+    "UserId": "",
+    "Token": ""
+  },
+  "VoiceChat": {
+    "AppId": "",
+    "RoomId": "",
+    "TaskId": "",
+    "AgentConfig": {
+      "TargetUserId": [
+        ""
+      ],
+      "WelcomeMessage": "你好，我是小宁，有什么需要帮忙的吗？",
+      "UserId": "",
+      "EnableConversationStateCallback": true
+    },
+    "Config": {
+      "ASRConfig": {
+        "Provider": "volcano",
+        "ProviderParams": {
+          "Mode": "smallmodel",
+          "AppId": "",
+          "Cluster": "volcengine_streaming_common"
+        }
+      },
+      "TTSConfig": {
+        "Provider": "volcano",
+        "ProviderParams": {
+          "app": {
+            "appid": "",
+            "cluster": "volcano_tts"
+          },
+          "audio": {
+            "voice_type": "BV001_streaming",
+            "speed_ratio": 1,
+            "pitch_ratio": 1,
+            "volume_ratio": 1
+          }
+        }
+      },
+      "LLMConfig": {
+        "Mode": "ArkV3",
+        "EndPointId": "",
+        "SystemMessages": [
+          "你是小宁，性格幽默又善解人意。你在表达时需简明扼要，有自己的观点。"
+        ],
+        "VisionConfig": {
+          "Enable": false
+        }
+      },
+      "AvatarConfig": {
+        "Enabled": false,
+        "AvatarType": "3min",
+        "AvatarRole": "250623-zhibo-linyunzhi",
+        "BackgroundUrl": "",
+        "VideoBitrate": 2000,
+        "AvatarAppID": "",
+        "AvatarToken": ""
+      },
+      "InterruptMode": 0
+    }
+  }
+}

backend/signer.py

@@ -0,0 +1,112 @@
+"""
+Copyright 2025 Beijing Volcano Engine Technology Co., Ltd. All Rights Reserved.
+SPDX-license-identifier: BSD-3-Clause
+
+Migrated from @volcengine/openapi Signer (AWS SigV4 compatible)
+Reference: https://www.volcengine.com/docs/6348/69828
+"""
+import hashlib
+import hmac
+import json
+from datetime import datetime, timezone
+from urllib.parse import quote
+
+
+def _sha256_hex(data: bytes) -> str:
+    return hashlib.sha256(data).hexdigest()
+
+
+def _hmac_sha256(key: bytes, data: str) -> bytes:
+    return hmac.new(key, data.encode("utf-8"), hashlib.sha256).digest()
+
+
+def _get_signing_key(secret_key: str, date_str: str, region: str, service: str) -> bytes:
+    k_date = _hmac_sha256(("HMAC-SHA256" + secret_key).encode("utf-8"), date_str)
+    k_region = _hmac_sha256(k_date, region)
+    k_service = _hmac_sha256(k_region, service)
+    k_signing = _hmac_sha256(k_service, "request")
+    return k_signing
+
+
+class Signer:
+    """
+    Signs requests to Volcengine OpenAPI using AWS SigV4-compatible signing.
+    """
+
+    def __init__(self, request_data: dict, service: str):
+        """
+        request_data: {
+            region: str,
+            method: str,
+            params: dict,      # query params (Action, Version, ...)
+            headers: dict,
+            body: dict,
+        }
+        service: e.g. "rtc"
+        """
+        self.region = request_data.get("region", "cn-north-1")
+        self.method = request_data.get("method", "POST").upper()
+        self.params = request_data.get("params", {})
+        self.headers = request_data.get("headers", {})
+        self.body = request_data.get("body", {})
+        self.service = service
+
+    def add_authorization(self, account_config: dict):
+        """
+        Computes and injects Authorization + X-Date headers into self.headers.
+        account_config: { accessKeyId: str, secretKey: str }
+        """
+        access_key = account_config["accessKeyId"]
+        secret_key = account_config["secretKey"]
+
+        now = datetime.now(timezone.utc)
+        date_str = now.strftime("%Y%m%d")
+        datetime_str = now.strftime("%Y%m%dT%H%M%SZ")
+
+        self.headers["X-Date"] = datetime_str
+        self.headers["X-Content-Sha256"] = _sha256_hex(
+            json.dumps(self.body, separators=(",", ":"), ensure_ascii=False).encode("utf-8")
+        )
+
+        # Canonical headers: sorted lowercase header names
+        signed_header_names = sorted(k.lower() for k in self.headers)
+        canonical_headers = "".join(
+            f"{k}:{self.headers[next(h for h in self.headers if h.lower() == k)]}\n"
+            for k in signed_header_names
+        )
+        signed_headers_str = ";".join(signed_header_names)
+
+        # Canonical query string
+        sorted_params = sorted(self.params.items())
+        canonical_qs = "&".join(
+            f"{quote(str(k), safe='')}={quote(str(v), safe='')}"
+            for k, v in sorted_params
+        )
+
+        # Canonical request
+        body_hash = self.headers["X-Content-Sha256"]
+        canonical_request = "\n".join([
+            self.method,
+            "/",
+            canonical_qs,
+            canonical_headers,
+            signed_headers_str,
+            body_hash,
+        ])
+
+        credential_scope = f"{date_str}/{self.region}/{self.service}/request"
+        string_to_sign = "\n".join([
+            "HMAC-SHA256",
+            datetime_str,
+            credential_scope,
+            _sha256_hex(canonical_request.encode("utf-8")),
+        ])
+
+        signing_key = _get_signing_key(secret_key, date_str, self.region, self.service)
+        signature = hmac.new(signing_key, string_to_sign.encode("utf-8"), hashlib.sha256).hexdigest()
+
+        self.headers["Authorization"] = (
+            f"HMAC-SHA256 Credential={access_key}/{credential_scope}, "
+            f"SignedHeaders={signed_headers_str}, "
+            f"Signature={signature}"
+        )

backend/token.py

@@ -0,0 +1,102 @@
+"""
+Copyright 2025 Beijing Volcano Engine Technology Co., Ltd. All Rights Reserved.
+SPDX-license-identifier: BSD-3-Clause
+
+Migrated from Server/token.js
+"""
+import hashlib
+import hmac
+import random
+import struct
+import time
+import base64
+
+VERSION = "001"
+VERSION_LENGTH = 3
+APP_ID_LENGTH = 24
+
+_random_nonce = random.randint(0, 0xFFFFFFFF)
+
+privileges = {
+    "PrivPublishStream": 0,
+    "privPublishAudioStream": 1,
+    "privPublishVideoStream": 2,
+    "privPublishDataStream": 3,
+    "PrivSubscribeStream": 4,
+}
+
+
+class ByteBuf:
+    def __init__(self):
+        self._buf = bytearray()
+
+    def put_uint16(self, v: int) -> "ByteBuf":
+        self._buf += struct.pack("<H", v)
+        return self
+
+    def put_uint32(self, v: int) -> "ByteBuf":
+        self._buf += struct.pack("<I", v)
+        return self
+
+    def put_bytes(self, b: bytes) -> "ByteBuf":
+        self.put_uint16(len(b))
+        self._buf += b
+        return self
+
+    def put_string(self, s: str) -> "ByteBuf":
+        return self.put_bytes(s.encode("utf-8"))
+
+    def put_tree_map_uint32(self, m: dict) -> "ByteBuf":
+        if not m:
+            self.put_uint16(0)
+            return self
+        self.put_uint16(len(m))
+        for key, value in m.items():
+            self.put_uint16(int(key))
+            self.put_uint32(int(value))
+        return self
+
+    def pack(self) -> bytes:
+        return bytes(self._buf)
+
+
+def _encode_hmac(key: str, message: bytes) -> bytes:
+    return hmac.new(key.encode("utf-8"), message, hashlib.sha256).digest()
+
+
+class AccessToken:
+    def __init__(self, app_id: str, app_key: str, room_id: str, user_id: str):
+        self.app_id = app_id
+        self.app_key = app_key
+        self.room_id = room_id
+        self.user_id = user_id
+        self.issued_at = int(time.time())
+        self.nonce = _random_nonce
+        self.expire_at = 0
+        self._privileges: dict = {}
+
+    def add_privilege(self, privilege: int, expire_timestamp: int):
+        self._privileges[privilege] = expire_timestamp
+        if privilege == privileges["PrivPublishStream"]:
+            self._privileges[privileges["privPublishVideoStream"]] = expire_timestamp
+            self._privileges[privileges["privPublishAudioStream"]] = expire_timestamp
+            self._privileges[privileges["privPublishDataStream"]] = expire_timestamp
+
+    def expire_time(self, expire_timestamp: int):
+        self.expire_at = expire_timestamp
+
+    def _pack_msg(self) -> bytes:
+        buf = ByteBuf()
+        buf.put_uint32(self.nonce)
+        buf.put_uint32(self.issued_at)
+        buf.put_uint32(self.expire_at)
+        buf.put_string(self.room_id)
+        buf.put_string(self.user_id)
+        buf.put_tree_map_uint32(self._privileges)
+        return buf.pack()
+
+    def serialize(self) -> str:
+        msg = self._pack_msg()
+        signature = _encode_hmac(self.app_key, msg)
+        content = ByteBuf().put_bytes(msg).put_bytes(signature).pack()
+        return VERSION + self.app_id + base64.b64encode(content).decode("utf-8")