fix: 修改modal_infer.exe 为单独打包 + CI support

.github/workflows/build-release-conda.yml

@@ -648,4 +648,4 @@ jobs:
         repository: ${{ github.repository }}
         tag_name: ${{ github.ref }}
         files: faster_whisper_transwithai_windows_cu128-chickenrice.zip
-        token: ${{ secrets.GITHUB_TOKEN }}
+        token: ${{ secrets.GITHUB_TOKEN }}

build_windows.py

@@ -210,6 +210,43 @@ def build():
     # Verify build succeeded and check for CUDA libraries
     if result.returncode == 0:
         dist_dir = Path("dist/faster_whisper_transwithai_chickenrice")
+        # Build modal_infer if modal.spec is present (separate target).
+        modal_spec = Path("modal.spec")
+        if modal_spec.exists():
+            # Ensure modal dependencies are available in the current env.
+            try:
+                import modal  # noqa: F401
+                import questionary  # noqa: F401
+            except ImportError:
+                print("\nmodal/questionary not found; installing for modal.spec build...")
+                install_cmd = [
+                    sys.executable, "-m", "pip", "install",
+                    "modal", "questionary",
+                ]
+                install_result = subprocess.run(install_cmd, capture_output=False)
+                if install_result.returncode != 0:
+                    print("\nFailed to install modal/questionary.")
+                    return 1
+
+            modal_cmd = [
+                sys.executable, "-m", "PyInstaller",
+                "--clean",
+                "--noconfirm",
+                "--distpath", str(Path("dist") / "faster_whisper_transwithai_chickenrice"),
+                "--workpath", str(Path("build") / "modal"),
+                str(modal_spec),
+            ]
+            print(f"\nRunning: {' '.join(modal_cmd)}")
+            modal_result = subprocess.run(modal_cmd, capture_output=False)
+            if modal_result.returncode != 0:
+                print("\nModal build failed!")
+                return 1
+
+        dist_root = Path("dist")
+        dist_dir = dist_root / "faster_whisper_transwithai_chickenrice"
+        engine_dir = dist_root / "engine"
+        client_dir = dist_root / "client"
+
         if dist_dir.exists():
             # Quick verification of critical libraries
             print("\nVerifying CUDA libraries in distribution...")
@@ -258,4 +295,5 @@ def build():
     return 0
 
 if __name__ == "__main__":
-    sys.exit(build())
+    sys.exit(build())
+

environment-modal.yml

@@ -0,0 +1,21 @@
+# Conda environment for Modal inference (local client only)
+# This environment is for running modal_infer.py locally to submit jobs to Modal
+name: faster-whisper-modal
+channels:
+  - conda-forge
+  - defaults
+
+dependencies:
+  # Python version
+  - python=3.10
+
+  # Core dependencies
+  - pip
+
+  # Pip dependencies
+  - pip:
+      # Modal client for submitting jobs
+      - modal
+
+      # Interactive CLI prompts
+      - questionary

modal.spec

@@ -0,0 +1,54 @@
+# -*- mode: python ; coding: utf-8 -*-
+import os
+from PyInstaller.utils.hooks import collect_all
+
+block_cipher = None
+
+datas = [("environment-cuda128.yml", ".")]
+binaries = []
+hiddenimports = []
+
+for package in ["modal", "questionary", "prompt_toolkit", "rich", "typer", "click"]:
+    try:
+        pkg_datas, pkg_binaries, pkg_hiddenimports = collect_all(package)
+        datas += pkg_datas
+        binaries += pkg_binaries
+        hiddenimports += pkg_hiddenimports
+    except Exception:
+        pass
+
+a = Analysis(
+    ["modal_infer.py"],
+    pathex=[],
+    binaries=binaries,
+    datas=datas,
+    hiddenimports=hiddenimports,
+    hookspath=[],
+    hooksconfig={},
+    runtime_hooks=[],
+    excludes=[],
+    noarchive=False,
+)
+
+pyz = PYZ(a.pure, a.zipped_data, cipher=block_cipher)
+
+exe = EXE(
+    pyz,
+    a.scripts,
+    a.binaries,
+    a.zipfiles,
+    a.datas,
+    exclude_binaries=False,
+    name="modal_infer",
+    debug=False,
+    bootloader_ignore_signals=False,
+    strip=False,
+    upx=False,
+    console=True,
+    disable_windowed_traceback=False,
+    argv_emulation=False,
+    target_arch=None,
+    codesign_identity=None,
+    entitlements_file=None,
+    icon="transwithai.ico" if os.path.exists("transwithai.ico") else None,
+)

modal_infer.py

@@ -1,17 +1,38 @@
-"""feature-modal: 交互式 CLI，完成 Modal App 构建、音频上传、推理执行与结果回传。"""
-
 from __future__ import annotations
 
 import argparse
+import io
 import logging
 import sys
 from dataclasses import dataclass
 from datetime import datetime
-from pathlib import Path
+from pathlib import Path, PurePosixPath
 import subprocess
 from typing import Dict, List, Optional, Sequence, Tuple
 from uuid import uuid4
 
+def ensure_utf8_stdio() -> None:
+    for name in ("stdout", "stderr"):
+        stream = getattr(sys, name, None)
+        if stream is None:
+            continue
+        try:
+            encoding = getattr(stream, "encoding", None)
+            if encoding and encoding.lower().startswith("utf-8"):
+                continue
+            if hasattr(stream, "reconfigure"):
+                stream.reconfigure(encoding="utf-8", errors="replace")
+            elif hasattr(stream, "buffer"):
+                setattr(
+                    sys,
+                    name,
+                    io.TextIOWrapper(stream.buffer, encoding="utf-8", errors="replace"),
+                )
+        except Exception:
+            pass
+
+ensure_utf8_stdio()
+
 try:
     import questionary  # type: ignore
     from questionary import Choice  # type: ignore
@@ -28,11 +49,11 @@
 APP_NAME = "Faster-Whisper-TransWithAI-ChickenRice"
 REPO_URL = "https://github.com/TransWithAI/Faster-Whisper-TransWithAI-ChickenRice"
 VOLUME_NAME = "Faster_Whisper"
-VOLUME_ROOT = Path("/Faster_Whisper")
+VOLUME_ROOT = "/Faster_Whisper"
 REMOTE_MOUNT = VOLUME_ROOT
-APP_ROOT_REL = Path(APP_NAME)
-SESSION_SUBDIR = Path("sessions")
-REPO_VOLUME_DIR = VOLUME_ROOT / "repo"
+APP_ROOT_REL = APP_NAME
+SESSION_SUBDIR = "sessions"
+REPO_VOLUME_DIR = f"{VOLUME_ROOT}/repo"
 SUB_FORMATS = "srt,vtt,lrc"
 SUB_SUFFIXES = {".srt", ".vtt", ".lrc"}
 AUDIO_SUFFIXES = {
@@ -63,6 +84,10 @@
     "B200",
 ]
 
+def resolve_resource_path(filename: str) -> Path:
+    base_dir = Path(getattr(sys, "_MEIPASS", Path(__file__).resolve().parent))
+    return base_dir / filename
+
 
 @dataclass
 class ModelProfile:
@@ -117,7 +142,8 @@ def rel_to_volume_path(path: Path) -> str:
 
 
 def rel_to_container_path(path: Path) -> str:
-    return str((REMOTE_MOUNT / path).as_posix())
+    base = PurePosixPath(REMOTE_MOUNT)
+    return str((base / path.as_posix()).as_posix())
 
 
 def volume_path_to_relative(path: str) -> Path:
@@ -321,7 +347,7 @@ def upload_single_file(
         base_dir: 基础目录（用于文件夹模式，输出到此目录）
     """
     session_id = f"{datetime.now().strftime('%Y%m%d-%H%M%S')}-{uuid4().hex[:6]}"
-    remote_session_rel = SESSION_SUBDIR / session_id
+    remote_session_rel = Path(SESSION_SUBDIR) / session_id
     remote_logs_rel = remote_session_rel / "logs"
 
     # 使用固定文件名避免全角字符等问题
@@ -389,7 +415,7 @@ def build_modal_image() -> modal.Image:
         modal.Image.micromamba(python_version="3.10")
         .apt_install("git")
         .micromamba_install(
-            spec_file="environment-cuda128.yml",
+            spec_file=str(resolve_resource_path("environment-cuda128.yml")),
             channels=["conda-forge", "defaults"],
         )
         .pip_install("modal", "questionary")
@@ -625,7 +651,7 @@ def run(cmd: Sequence[str], cwd: Optional[str] = None, env: Optional[dict] = Non
         subprocess.run(cmd, check=True, cwd=cwd, env=env)
 
     mount_root = Path(job["mount_root"])
-    repo_dir = REPO_VOLUME_DIR
+    repo_dir = Path(REPO_VOLUME_DIR)
 
     # log 文件放在 session 目录下，而不是 logs 子目录
     session_dir = Path(job["remote_output_dir"])
@@ -640,10 +666,10 @@ def log(msg: str) -> None:
 
     if not (repo_dir / ".git").exists():
         log("开始克隆仓库...")
-        run(["git", "clone", REPO_URL, str(repo_dir)])
+        run(["git", "clone", "--depth", "1", REPO_URL, str(repo_dir)])
     else:
         log("更新仓库...")
-        run(["git", "-C", str(repo_dir), "fetch", "origin", "main"])
+        run(["git", "-C", str(repo_dir), "fetch", "origin"])
         run(["git", "-C", str(repo_dir), "reset", "--hard", "origin/main"])
 
     model_profile = job["model_profile"]
@@ -689,16 +715,12 @@ def snapshot(path: str) -> set:
     cmd = [
         "python",
         str(repo_dir / "infer.py"),
-        "--device",
-        "cuda",
-        "--model_name_or_path",
-        str(model_path),
-        "--sub_formats",
-        job["sub_formats"],
-        "--log_level",
-        "INFO",
-        "--output_dir",
-        str(output_dir),
+        "--audio_suffixes", "mp3,wav,flac,m4a,aac,ogg,wma,mp4,mkv,avi,mov,webm,flv,wmv",
+        "--device","cuda",
+        "--model_name_or_path",str(model_path),
+        "--sub_formats",job["sub_formats"],
+        "--log_level","INFO",
+        "--output_dir",str(output_dir),
     ]
     if job["enable_batching"]:
         cmd.append("--enable_batching")

project.spec

@@ -346,7 +346,7 @@ datas += [
 ]
 
 a = Analysis(
-    ['infer.py', 'modal_infer.py'],
+    ['infer.py'],
     pathex=[],
     binaries=binaries,
     datas=datas,
@@ -374,9 +374,9 @@ a = Analysis(
 
 pyz = PYZ(a.pure, a.zipped_data, cipher=block_cipher)
 
-infer_exe = EXE(
+exe = EXE(
     pyz,
-    [a.scripts[0]],
+    a.scripts,
     [],
     exclude_binaries=True,
     name='infer',
@@ -393,33 +393,13 @@ infer_exe = EXE(
     icon='transwithai.ico' if os.path.exists('transwithai.ico') else None,
 )
 
-modal_exe = EXE(
-    pyz,
-    [a.scripts[1]],
-    [],
-    exclude_binaries=True,
-    name='modal_infer',
-    debug=False,
-    bootloader_ignore_signals=False,
-    strip=False,
-    upx=False,
-    console=True,
-    disable_windowed_traceback=False,
-    argv_emulation=False,
-    target_arch=None,
-    codesign_identity=None,
-    entitlements_file=None,
-    icon='transwithai.ico' if os.path.exists('transwithai.ico') else None,
-)
-
 coll = COLLECT(
-    infer_exe,
-    modal_exe,
+    exe,
     a.binaries,
     a.zipfiles,
     a.datas,
     strip=False,
     upx=False,
     upx_exclude=[],
     name='faster_whisper_transwithai_chickenrice',
-)
+)

使用说明.txt

@@ -100,13 +100,14 @@ GPU模式（仅限NVIDIA显卡）:
 
 1. 环境配置：
 
-使用现有的 Conda 环境（已包含 modal 支持）：
+使用 Conda 创建轻量级环境（仅需 modal 和 questionary 库）：
 
 ```bash
-conda activate faster-whisper-cu118  # 或 cu122, cu128
+conda env create -f environment-modal.yml
+conda activate faster-whisper-modal
 ```
 
-或在现有环境中手动安装：
+或手动安装：
 ```bash
 pip install modal questionary
 ```