diff --git a/bin/api/__init__.py b/bin/api/__init__.py new file mode 100644 index 0000000..418d8ca --- /dev/null +++ b/bin/api/__init__.py @@ -0,0 +1,6 @@ +"""Pure API clients for external services (no CLI deps).""" + +from .image import generate_image +from .voice import synthesize_tts + +__all__ = ["generate_image", "synthesize_tts"] diff --git a/bin/api/image.py b/bin/api/image.py new file mode 100644 index 0000000..7007bab --- /dev/null +++ b/bin/api/image.py @@ -0,0 +1,57 @@ +"""Image generation API client (Doubao/Seedream).""" + +import base64 +import os +import urllib.request +from pathlib import Path + +_RATIO_MAP = { + "1:1": "1024x1024", + "4:3": "1024x768", + "3:4": "768x1024", + "16:9": "1280x720", + "9:16": "720x1280", +} + + +def generate_image( + api_key: str, + prompt: str, + size: str | None = None, + ratio: str | None = None, + output: str | None = None, + watermark: bool = True, + b64: bool = False, +) -> str: + """Generate image using Doubao/Seedream API. + + Returns absolute file path if saved, or URL string if output is None. + """ + from openai import OpenAI + + resolved_size = size or (ratio and _RATIO_MAP.get(ratio)) or "2K" + + client = OpenAI( + base_url="https://ark.cn-beijing.volces.com/api/v3", + api_key=api_key, + ) + + resp = client.images.generate( + model="doubao-seedream-5-0-260128", + prompt=prompt, + size=resolved_size, + response_format="b64_json" if b64 else "url", + extra_body={"watermark": watermark}, + ) + + if b64: + image_bytes = base64.b64decode(resp.data[0].b64_json) + save_path = output or "generated_image.png" + Path(save_path).write_bytes(image_bytes) + return os.path.abspath(save_path) + else: + url = resp.data[0].url + if output: + urllib.request.urlretrieve(url, output) + return os.path.abspath(output) + return url diff --git a/bin/api/voice.py b/bin/api/voice.py new file mode 100644 index 0000000..8be5a49 --- /dev/null +++ b/bin/api/voice.py @@ -0,0 +1,93 @@ +"""Voice / TTS API client via Volcengine (ByteDance) WebSocket API.""" + +import asyncio +import json +import uuid +from pathlib import Path + +from bin.commands.protocols import EventType, MsgType, full_client_request, receive_message + +_RESOURCE_ID = "volc.service_type.10029" +_ENDPOINT = "wss://openspeech.bytedance.com/api/v3/tts/unidirectional/stream" + +_VOICE_CHOICES = [ + "zh_female_cancan_mars_bigtts", + "zh_female_shuangkuaisisi_moon_bigtts", + "zh_male_wennuanahu_moon_bigtts", + "zh_male_sunwukong_moon_bigtts", +] + +_FORMAT_CHOICES = ["mp3", "wav", "pcm"] + + +def synthesize_tts( + appid: str, + token: str, + text: str, + speaker: str = "zh_male_wennuanahu_moon_bigtts", + fmt: str = "mp3", + speech_rate: int = 0, + output_path: str = "output.mp3", +) -> str: + """Synchronous TTS synthesis via Volcengine WebSocket API. + + Returns absolute path to the output audio file. + """ + try: + import websockets + except ImportError as e: + raise RuntimeError(f"Missing dependency: {e}. Run: pip install websockets") from e + + headers = { + "X-Api-App-Id": appid, + "X-Api-Access-Key": token, + "X-Api-Resource-Id": _RESOURCE_ID, + "X-Api-Connect-Id": str(uuid.uuid4()), + } + + async def _run() -> bytearray: + ws = await websockets.connect( + _ENDPOINT, + additional_headers=headers, + max_size=10 * 1024 * 1024, + open_timeout=30, + ) + + request = { + "user": {"uid": str(uuid.uuid4())}, + "req_params": { + "speaker": speaker, + "audio_params": { + "format": fmt, + "sample_rate": 24000, + "speech_rate": speech_rate, + }, + "text": text, + }, + } + + await full_client_request(ws, json.dumps(request).encode()) + + audio_data = bytearray() + while True: + msg = await receive_message(ws) + if msg.type == MsgType.FullServerResponse and msg.event == EventType.SessionFinished: + break + elif msg.type == MsgType.AudioOnlyServer: + audio_data.extend(msg.payload) + elif msg.type == MsgType.Error: + await ws.close() + raise RuntimeError( + f"TTS service error: {msg.payload.decode('utf-8', errors='replace')}" + ) + + await ws.close() + return audio_data + + audio_data = asyncio.run(_run()) + if not audio_data: + raise RuntimeError("No audio data received from TTS service") + + out = Path(output_path) + out.write_bytes(audio_data) + return str(out.resolve()) diff --git a/bin/commands/image.py b/bin/commands/image.py index 5cb4aa9..ba6689e 100644 --- a/bin/commands/image.py +++ b/bin/commands/image.py @@ -1,13 +1,11 @@ """Image utilities.""" -import base64 -import os import subprocess -import urllib.request from pathlib import Path import click +from bin.api.image import _RATIO_MAP, generate_image from bin.config import config from bin.utils import handle_errors, run_command @@ -187,15 +185,6 @@ def compress_images(files, target_mb, max_width, quality, suffix, dry_run): click.echo(f"Compressed: {img_file} -> {output} ({size_mb:.2f} MB)") -_RATIO_MAP = { - "1:1": "1024x1024", - "4:3": "1024x768", - "3:4": "768x1024", - "16:9": "1280x720", - "9:16": "720x1280", -} - - @image.command("generate") @click.argument("prompt", required=False) @click.option("-f", "--file", type=click.Path(exists=True), help="从文件读取 prompt") @@ -207,8 +196,6 @@ _RATIO_MAP = { @handle_errors def generate_image_cmd(prompt, file, size, ratio, output, no_watermark, b64): """使用豆包/Seedream 模型生成图片。""" - from openai import OpenAI - if file: prompt = Path(file).read_text(encoding="utf-8").strip() elif not prompt: @@ -222,30 +209,13 @@ def generate_image_cmd(prompt, file, size, ratio, output, no_watermark, b64): if size and ratio: click.echo("Warning: --size 和 --ratio 同时指定，--size 优先", err=True) - resolved_size = size or (ratio and _RATIO_MAP[ratio]) or "2K" - - client = OpenAI( - base_url="https://ark.cn-beijing.volces.com/api/v3", + result = generate_image( api_key=api_key, - ) - - resp = client.images.generate( - model="doubao-seedream-5-0-260128", prompt=prompt, - size=resolved_size, - response_format="b64_json" if b64 else "url", - extra_body={"watermark": not no_watermark}, + size=size, + ratio=ratio, + output=output, + watermark=not no_watermark, + b64=b64, ) - - if b64: - image_bytes = base64.b64decode(resp.data[0].b64_json) - save_path = output or "generated_image.png" - Path(save_path).write_bytes(image_bytes) - click.echo(os.path.abspath(save_path)) - else: - url = resp.data[0].url - if output: - urllib.request.urlretrieve(url, output) - click.echo(os.path.abspath(output)) - else: - click.echo(url) + click.echo(result) diff --git a/bin/commands/voice.py b/bin/commands/voice.py index 6e00611..b399fe9 100644 --- a/bin/commands/voice.py +++ b/bin/commands/voice.py @@ -1,19 +1,14 @@ """Voice / TTS utilities via Volcengine (ByteDance) WebSocket API.""" -import asyncio -import json import os -import uuid from pathlib import Path import click +from bin.api.voice import synthesize_tts from bin.config import config from bin.utils import handle_errors -_RESOURCE_ID = "volc.service_type.10029" -_ENDPOINT = "wss://openspeech.bytedance.com/api/v3/tts/unidirectional/stream" - _VOICE_CHOICES = [ "zh_female_cancan_mars_bigtts", "zh_female_shuangkuaisisi_moon_bigtts", @@ -41,17 +36,6 @@ def _synthesize_sync( output_path: str, ) -> str: """Synchronous wrapper around async WebSocket TTS.""" - try: - from bin.commands.protocols import ( - EventType, - MsgType, - full_client_request, - receive_message, - ) - import websockets - except ImportError as e: - raise RuntimeError(f"Missing dependency: {e}. Run: pip install websockets") - appid, access_token = _get_credentials() if not appid or not access_token: raise click.UsageError( @@ -60,61 +44,15 @@ def _synthesize_sync( " mytoolkit env set volc_access_token " ) - headers = { - "X-Api-App-Id": appid, - "X-Api-Access-Key": access_token, - "X-Api-Resource-Id": _RESOURCE_ID, - "X-Api-Connect-Id": str(uuid.uuid4()), - } - - async def _run() -> bytearray: - try: - ws = await websockets.connect( - _ENDPOINT, - additional_headers=headers, - max_size=10 * 1024 * 1024, - open_timeout=30, - ) - except Exception as e: - raise RuntimeError(f"WebSocket connection failed: {e}") from e - - request = { - "user": {"uid": str(uuid.uuid4())}, - "req_params": { - "speaker": speaker, - "audio_params": { - "format": fmt, - "sample_rate": 24000, - "speech_rate": speech_rate, - }, - "text": text, - }, - } - - await full_client_request(ws, json.dumps(request).encode()) - - audio_data = bytearray() - while True: - msg = await receive_message(ws) - if msg.type == MsgType.FullServerResponse and msg.event == EventType.SessionFinished: - break - elif msg.type == MsgType.AudioOnlyServer: - audio_data.extend(msg.payload) - elif msg.type == MsgType.Error: - await ws.close() - raise RuntimeError( - f"TTS service error: {msg.payload.decode('utf-8', errors='replace')}" - ) - - await ws.close() - return audio_data - - audio_data = asyncio.run(_run()) - if not audio_data: - raise RuntimeError("No audio data received from TTS service") - - Path(output_path).write_bytes(audio_data) - return output_path + return synthesize_tts( + appid=appid, + token=access_token, + text=text, + speaker=speaker, + fmt=fmt, + speech_rate=speech_rate, + output_path=output_path, + ) @click.group() diff --git a/templates/md-to-docx/default/template.docx b/templates/md-to-docx/default/template.docx index 7506cb8..c30e09b 100644 Binary files a/templates/md-to-docx/default/template.docx and b/templates/md-to-docx/default/template.docx differ