208 lines
5.4 KiB
Markdown
208 lines
5.4 KiB
Markdown
# SmartMeeting 智能会议
|
||
|
||
这是一个基于 Docker Compose 的智能会议部署项目，包含 Speakr 前后端、Paraformer ASR、声纹向量提取服务，以及实时 WebSocket 转写桥接服务。
|
||
|
||
## 目录结构
|
||
|
||
```text
|
||
/storage01/home/xll/202607/voice
|
||
├── docker-compose.yml # 统一启动文件
|
||
├── config.env # 运行环境变量，主要改这里
|
||
├── asr/main.py # ASR 适配服务代码
|
||
├── speakr/ # Speakr 前后端代码
|
||
├── voice-dected/ # 声纹向量提取服务代码
|
||
├── fastapi_wss/ # 实时 WebSocket 桥接服务代码
|
||
└── deploy/data/ # 运行数据，已被 git 忽略
|
||
```
|
||
|
||
## 启动和停止
|
||
|
||
启动全部服务：
|
||
|
||
```bash
|
||
cd /storage01/home/xll/202607/voice
|
||
docker compose up -d
|
||
```
|
||
|
||
查看状态和日志：
|
||
|
||
```bash
|
||
docker compose ps
|
||
docker compose logs -f
|
||
```
|
||
|
||
停止服务：
|
||
|
||
```bash
|
||
docker compose down
|
||
```
|
||
|
||
## 服务和端口
|
||
|
||
| 服务 | 容器名 | 端口 | 作用 |
|
||
| --- | --- | --- | --- |
|
||
| Speakr | `voice-speakr` | `8899` | 页面和后端 |
|
||
| ASR | `asr-test` | `59805` | Paraformer HTTP / WebSocket 识别 |
|
||
| Python 实时 ASR | `funasr-online` | `10096` | FunASR Python WebSocket，保留用于回退 |
|
||
| 官方 Runtime 实时 ASR | `funasr-runtime-2pass` | `10097` | FunASR 官方 C++/ONNX 2pass WebSocket |
|
||
| 声纹提取 | `voice-dected` | `18000` | `/extract_embedding` 声纹向量服务 |
|
||
| 实时桥接 | `fastapi-wss` | `10095` | 页面实时 ASR WebSocket 桥接 |
|
||
| Redis | `voice-redis` | `16380` | 实时桥接的缓存/会话存储 |
|
||
|
||
页面访问地址：
|
||
|
||
```text
|
||
http://192.168.0.46:8899/tool/speakr/
|
||
```
|
||
|
||
浏览器麦克风录音需要安全上下文。局域网 IP 的 HTTP 页面可能无法调用麦克风。最快的本机访问方式是开 SSH 隧道：
|
||
|
||
```bash
|
||
ssh -L 8899:127.0.0.1:8899 xll@192.168.0.46
|
||
```
|
||
|
||
然后浏览器打开：
|
||
|
||
```text
|
||
http://localhost:8899/tool/speakr/
|
||
```
|
||
|
||
## 配置文件
|
||
|
||
主要修改这个文件：
|
||
|
||
```text
|
||
/storage01/home/xll/202607/voice/config.env
|
||
```
|
||
|
||
改服务地址、模型地址、LLM 配置、声纹匹配阈值时，优先改 `config.env`。修改后执行：
|
||
|
||
```bash
|
||
cd /storage01/home/xll/202607/voice
|
||
docker compose up -d
|
||
```
|
||
|
||
重要变量：
|
||
|
||
```env
|
||
ASR_BASE_URL=http://asr-test:59805
|
||
FUNASR_WEBSOCKET_IP=ws://fastapi-wss:10095/websocket_offline
|
||
TARGET_WS_URL=ws://funasr-runtime-2pass:10095
|
||
# 如需回退到 Python 版实时 ASR，可改为：
|
||
# TARGET_WS_URL=ws://funasr-online:10096
|
||
VOICE_DETECTED_URL=http://voice-dected:8000
|
||
SPEAKER_DET_URL=http://voice-dected:8000/extract_embedding
|
||
REDIS_URL=redis://redis:6379/0
|
||
|
||
TEXT_MODEL_BASE_URL=http://192.168.0.46:59800/v1
|
||
TEXT_MODEL_NAME=46-qwen3.5-35B
|
||
TEXT_MODEL_API_KEY=none
|
||
|
||
VOICEPRINT_STORE=/data/asr_voiceprints.json
|
||
VOICEPRINT_MATCH_THRESHOLD=0.45
|
||
REALTIME_VOICEPRINT_ENABLED=true
|
||
REALTIME_VOICEPRINT_WINDOW_SECONDS=2.0
|
||
```
|
||
|
||
说明：
|
||
|
||
- Compose 内部服务互相访问时，用 `asr-test`、`voice-dected`、`redis` 这种服务名。
|
||
- 宿主机或外部机器访问时，用 `192.168.0.46:59805`、`192.168.0.46:18000`、`192.168.0.46:10095` 这种宿主机端口。
|
||
- `FUNASR_WEBSOCKET_IP` 是 Speakr 后端实时录音代理要连接的 WebSocket 地址。
|
||
- `TARGET_WS_URL` 是 `fastapi-wss` 再往后连接的 ASR WebSocket 地址。
|
||
- `VOICEPRINT_MATCH_THRESHOLD` 是声纹匹配阈值。值越低越容易匹配，值越高越不容易误匹配。
|
||
- `REALTIME_VOICEPRINT_WINDOW_SECONDS` 是实时声纹匹配窗口，当前为 2 秒；识别抖动时可调到 3-5 秒。
|
||
|
||
## 运行数据
|
||
|
||
运行数据放在：
|
||
|
||
```text
|
||
/storage01/home/xll/202607/voice/deploy/data/
|
||
```
|
||
|
||
常见子目录：
|
||
|
||
```text
|
||
deploy/data/speakr/uploads/ # 上传和录音文件
|
||
deploy/data/speakr/instance/ # Speakr SQLite 数据库
|
||
deploy/data/asr/ # ASR 声纹库
|
||
deploy/data/redis/ # Redis 持久化数据
|
||
```
|
||
|
||
这些运行数据已被 `.gitignore` 忽略，不会提交到仓库。
|
||
|
||
## 声纹流程
|
||
|
||
Speakr 声纹管理页面会调用 ASR 适配层的兼容接口：
|
||
|
||
```text
|
||
POST /voice_insert
|
||
GET /get_voice_name
|
||
DELETE /delete_voice
|
||
```
|
||
|
||
ASR 适配层会把声纹向量保存到：
|
||
|
||
```text
|
||
deploy/data/asr/asr_voiceprints.json
|
||
```
|
||
|
||
转录时，`/asr` 会先把上传音频转成 `16k 单声道 wav`，再调用 `voice-dected` 提取声纹向量，并和已录入声纹做相似度匹配。如果分数超过 `VOICEPRINT_MATCH_THRESHOLD`，返回匹配到的说话人名称。
|
||
|
||
目前这个逻辑是“整段音频匹配一个最像的人”。如果一段录音里多人轮流说话，还需要进一步接入分段说话人识别。
|
||
|
||
## 常用检查命令
|
||
|
||
检查 ASR：
|
||
|
||
```bash
|
||
curl http://127.0.0.1:59805/health
|
||
```
|
||
|
||
检查声纹列表：
|
||
|
||
```bash
|
||
curl http://127.0.0.1:59805/get_voice_name
|
||
```
|
||
|
||
检查 Speakr 配置：
|
||
|
||
```bash
|
||
curl http://127.0.0.1:8899/tool/speakr/api/config
|
||
```
|
||
|
||
检查实时 WebSocket 链路：
|
||
|
||
```bash
|
||
docker exec -it fastapi-wss python - <<'PY'
|
||
import asyncio, json, websockets
|
||
|
||
async def main():
|
||
async with websockets.connect("ws://127.0.0.1:10095") as ws:
|
||
await ws.send(json.dumps({"mode": "2pass", "is_speaking": True}))
|
||
await ws.send(json.dumps({"type": "stop", "is_speaking": False}))
|
||
print(await ws.recv())
|
||
|
||
asyncio.run(main())
|
||
PY
|
||
```
|
||
|
||
## Git 操作
|
||
|
||
仓库地址：
|
||
|
||
```text
|
||
https://gitea.zkzdht.com/Ascend/SmartMeeting.git
|
||
```
|
||
|
||
常用命令：
|
||
|
||
```bash
|
||
cd /storage01/home/xll/202607/voice
|
||
git status
|
||
git add .
|
||
git commit -m "更新说明"
|
||
git push
|
||
```
|