WorkBuddy
df56c2c137
内容分四块： 1、产品规划产出 —— MCN 短视频整合营销工作台的①段五份（1a 需求／1b 竞品／1c 画像／1d 策略／1e 场景）、②段两份（2a 功能／2b 布局）、③段界面（DESIGN.md 契约与令牌表 ＋ mcn-workbench.html 原型 ＋ 实测/会诊/审查三份 ＋ 23 张闸门截图）。 2、开源竞品调研 —— 5 个内容工作台项目的取证原始件与 1b 系列分析文档。 3、参考资料 —— 竞品视频抽帧 1145 张 ＋ 2 个源视频 ＋ 功能点截图。 4、机制侧 —— 协作脚本与状态台账、工作区记忆日志、抽帧/OCR 脚本。 .gitignore 只排运行时日志、脚本备份副本与一次性探针输出，其余按原样入库。
1.3 KiB
FunASR / SenseVoice transcription
OpenCreator can use a user-managed FunASR server through its OpenAI-compatible API. OpenCreator does not package Python, CUDA, model files, or a resident Python process.
Start the official server separately, for example:
pip install funasr fastapi uvicorn python-multipart
funasr-server --host 127.0.0.1 --model sensevoice --device cpu
In AI services → Transcription, choose FunASR / SenseVoice, set the server Base URL (normally http://127.0.0.1:8000/v1) and model alias (sensevoice, paraformer, or another alias exposed by the server). API Key is optional and is only sent as a Bearer credential when configured. The timeout applies to the transcription request.
The connection test checks /v1/models for the selected model; it does not execute a transcription or verify output quality. Transcription uses POST /v1/audio/transcriptions with multipart file, model, and response_format=verbose_json; the response is normalized into the runtime transcription contract. HTTP errors, invalid responses, timeouts, and cancellation remain executor failures.
See the official FunASR OpenAI-compatible API example for server setup and model details.