in recent work,my leader ai clip task assigened to me, this is a arduous task , i search web , programme Few plans , gpt tell me can take a look a name is video-use project , with an attitude of experimentation , it is meet my needs , I tried several programme , result is not bad ! easy say, this project is a codex/claude other agent used skill , it core is call ffmpeg、ffprobe、HyperFrames、Remotion、Manim、PIL/Pillow、ElevenLabs…….，What is listed here are all the tools and technologies that I have used in my projects ：
- ffmpeg、ffprobe
- HyperFrames
- Minimax Synchronous speech synthesis skill (this is my encapsulation a skill，used to request minimax API)
- get-bgm-for-mixkit (this is my encapsulation a crawl Mixkit websit get bgm skill)
- s3-rustfs-storage (this is my encapsulation a skill，used to request rustfs request API)
With these things, it can satisfy the leader , but i use codex , those colleagues who need to use the software unwilling codex ,even if it’s an application instead of a command-line interface (CLI) , and there is another big problem, skills frequent changes, if let them use codex application/cli , manage skills will be very troublesomeso, i think of a way docking webui
open-webui : this is a can dock gpt model or other compatible openai response model
kanna : can let webui bridge codex , but no have account system
facing this situation , i decide onself develop a bridge program , in open-webui allow Pipe , because of this, it was born codex-bridge-webui , if you interested , you can take a look , it is responsible for converting user messages, attachments, and conversation information from Open WebUI into tasks that Codex can execute, and maintaining separate working directories for each user and each conversation. During task execution, it returns status, logs, and results through SSE streams, allowing Open WebUI to use the local Codex Agent just like a regular chat model
so far, the requirements of the leader have been met