更新

server/extend/.gitignore

@@ -1,2 +1,4 @@
 *
-!.gitignore
+!.gitignore
+!whisper-asr/
+!whisper-asr/**

server/extend/whisper-asr/.dockerignore

@@ -0,0 +1,6 @@
+__pycache__/
+*.pyc
+.venv/
+venv/
+models/
+.env

server/extend/whisper-asr/.gitignore

@@ -0,0 +1,6 @@
+__pycache__/
+*.pyc
+.venv/
+venv/
+models/
+.env

server/extend/whisper-asr/Dockerfile

@@ -0,0 +1,26 @@
+FROM python:3.11-slim
+
+ENV PYTHONDONTWRITEBYTECODE=1 \
+    PYTHONUNBUFFERED=1 \
+    ASR_MODEL_CACHE=/models
+
+RUN apt-get update \
+    && apt-get install -y --no-install-recommends ca-certificates ffmpeg \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+COPY main.py .
+
+RUN useradd --create-home --shell /usr/sbin/nologin appuser \
+    && mkdir -p /models \
+    && chown -R appuser:appuser /app /models
+
+USER appuser
+
+EXPOSE 8791
+
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8791"]

server/extend/whisper-asr/README.md

@@ -0,0 +1,46 @@
+# Docker Whisper ASR
+
+本服务为私聊语音输入提供本地语音转文字能力。它只绑定服务器本机 `127.0.0.1:8791`，由 PHP 后端 `/api/chat/voiceToText` 代理访问，不直接暴露公网。
+
+## 启动
+
+```bash
+cd server/extend/whisper-asr
+docker compose up -d --build
+```
+
+首次启动会下载 Whisper 模型到 Docker volume `whisper_asr_models`，耗时取决于网络和模型大小。
+
+## 配置
+
+常用环境变量：
+
+| 变量 | 默认值 | 说明 |
+| --- | --- | --- |
+| `ASR_MODEL` | `small` | Whisper 模型名 |
+| `ASR_DEVICE` | `cpu` | `cpu` 或 `cuda` |
+| `ASR_COMPUTE_TYPE` | `int8` | CPU 默认建议 `int8` |
+| `ASR_LANGUAGE` | 空 | 留空自动识别，中文可设为 `zh` |
+| `ASR_MAX_FILE_MB` | `10` | 单文件最大大小 |
+| `ASR_API_TOKEN` | 空 | 配置后需要 Bearer Token |
+
+如果配置 `ASR_API_TOKEN`，需要同步更新后台 AI 配置 `whisper_asr_api_token`。
+
+## 验证
+
+```bash
+curl http://127.0.0.1:8791/health
+curl -F "file=@sample.m4a" http://127.0.0.1:8791/v1/transcribe
+```
+
+接口成功时返回：
+
+```json
+{
+  "text": "识别到的文字",
+  "language": "zh",
+  "language_probability": 0.9,
+  "duration": 3.2,
+  "duration_ms": 1200
+}
+```

server/extend/whisper-asr/docker-compose.yml

@@ -0,0 +1,28 @@
+services:
+  whisper-asr:
+    build: .
+    image: sport-era/whisper-asr:latest
+    container_name: sport-era-whisper-asr
+    restart: unless-stopped
+    environment:
+      ASR_MODEL: ${ASR_MODEL:-small}
+      ASR_DEVICE: ${ASR_DEVICE:-cpu}
+      ASR_COMPUTE_TYPE: ${ASR_COMPUTE_TYPE:-int8}
+      ASR_LANGUAGE: ${ASR_LANGUAGE:-}
+      ASR_BEAM_SIZE: ${ASR_BEAM_SIZE:-5}
+      ASR_MAX_FILE_MB: ${ASR_MAX_FILE_MB:-10}
+      ASR_API_TOKEN: ${ASR_API_TOKEN:-}
+      ASR_MODEL_CACHE: /models
+    ports:
+      - "127.0.0.1:8791:8791"
+    volumes:
+      - whisper_asr_models:/models
+    healthcheck:
+      test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://127.0.0.1:8791/health', timeout=3).read()"]
+      interval: 30s
+      timeout: 5s
+      retries: 3
+      start_period: 60s
+
+volumes:
+  whisper_asr_models:

server/extend/whisper-asr/main.py

@@ -0,0 +1,107 @@
+import os
+import tempfile
+import time
+from pathlib import Path
+from typing import Optional
+
+from fastapi import Depends, FastAPI, File, Header, HTTPException, UploadFile
+from faster_whisper import WhisperModel
+
+
+APP_NAME = "Sport Era Whisper ASR"
+ALLOWED_EXTENSIONS = {"mp3", "m4a", "aac", "wav", "amr", "mp4", "3gp"}
+
+app = FastAPI(title=APP_NAME)
+model: Optional[WhisperModel] = None
+
+
+def env_int(name: str, default: int) -> int:
+    try:
+        return int(os.getenv(name, str(default)))
+    except ValueError:
+        return default
+
+
+def get_model() -> WhisperModel:
+    global model
+    if model is None:
+        model_name = os.getenv("ASR_MODEL", "small")
+        device = os.getenv("ASR_DEVICE", "cpu")
+        compute_type = os.getenv("ASR_COMPUTE_TYPE", "int8")
+        cache_dir = os.getenv("ASR_MODEL_CACHE", "/models")
+        model = WhisperModel(
+            model_name,
+            device=device,
+            compute_type=compute_type,
+            download_root=cache_dir,
+        )
+    return model
+
+
+def verify_token(authorization: str = Header(default="")) -> None:
+    token = os.getenv("ASR_API_TOKEN", "").strip()
+    if not token:
+        return
+    expected = f"Bearer {token}"
+    if authorization != expected:
+        raise HTTPException(status_code=401, detail="Unauthorized")
+
+
+def file_extension(filename: str) -> str:
+    return Path(filename or "").suffix.lower().lstrip(".")
+
+
+@app.get("/health")
+def health() -> dict:
+    return {
+        "status": "ok",
+        "service": APP_NAME,
+        "model": os.getenv("ASR_MODEL", "small"),
+        "device": os.getenv("ASR_DEVICE", "cpu"),
+        "compute_type": os.getenv("ASR_COMPUTE_TYPE", "int8"),
+    }
+
+
+@app.post("/v1/transcribe")
+async def transcribe(file: UploadFile = File(...), _: None = Depends(verify_token)) -> dict:
+    extension = file_extension(file.filename)
+    if extension not in ALLOWED_EXTENSIONS:
+        raise HTTPException(status_code=400, detail="Unsupported audio format")
+
+    max_size_mb = max(1, env_int("ASR_MAX_FILE_MB", 10))
+    content = await file.read(max_size_mb * 1024 * 1024 + 1)
+    if not content:
+        raise HTTPException(status_code=400, detail="Empty audio file")
+    if len(content) > max_size_mb * 1024 * 1024:
+        raise HTTPException(status_code=413, detail=f"Audio file is larger than {max_size_mb}MB")
+
+    temp_path = ""
+    try:
+        with tempfile.NamedTemporaryFile(delete=False, suffix=f".{extension}") as temp_file:
+            temp_file.write(content)
+            temp_path = temp_file.name
+
+        language = os.getenv("ASR_LANGUAGE", "").strip() or None
+        beam_size = max(1, env_int("ASR_BEAM_SIZE", 5))
+        start_time = time.perf_counter()
+        segments, info = get_model().transcribe(
+            temp_path,
+            language=language,
+            beam_size=beam_size,
+            vad_filter=True,
+        )
+        text = "".join(segment.text for segment in segments).strip()
+        duration_ms = int((time.perf_counter() - start_time) * 1000)
+        return {
+            "text": text,
+            "language": info.language,
+            "language_probability": info.language_probability,
+            "duration": info.duration,
+            "duration_ms": duration_ms,
+        }
+    finally:
+        if temp_path:
+            try:
+                os.unlink(temp_path)
+            except OSError:
+                pass

server/extend/whisper-asr/requirements.txt

@@ -0,0 +1,4 @@
+fastapi>=0.115,<1.0
+uvicorn[standard]>=0.32,<1.0
+python-multipart>=0.0.20,<1.0
+faster-whisper>=1.1,<2.0