src

Demo: Multi-model CLI chat

Interactive command-line chat — same UX in three languages, all routing through one OpenAI-compatible endpoint.

What it does

• Type a message → streamed reply from the current model
• /model <id> — switch model on the fly (e.g. /model gpt-5, /model gemini-3.1-pro)
• /image <prompt> — generate an image with gpt-image-2
• /clear — clear chat history
• /exit — quit

The same commands and behavior work in all three implementations.

Run it

All versions read API_KEY and BASE_URL from environment variables. BASE_URL defaults to http://xdhdancer.top/v1 (the example gateway). Replace with any OpenAI-compatible endpoint to use your own provider.

Python

cd src/python
pip install -r requirements.txt
export API_KEY=sk-xxx
python chat.py

Node.js

cd src/node
npm install
export API_KEY=sk-xxx
node chat.js

Go

cd src/go
export API_KEY=sk-xxx
go mod tidy
go run chat.go

Sample session

connected to http://xdhdancer.top/v1
current model: claude-opus-4-7
type /exit to quit, /model <id> to switch, /image <prompt> to generate

> 用一句话解释 RAG
RAG 是检索增强生成,先从知识库检索相关内容再让 LLM 基于这些内容生成答案。

> /model gpt-5
(switched to gpt-5)

> 同样的问题再回答一次
RAG (Retrieval-Augmented Generation) combines a retriever and a generator: ...

> /image a cyberpunk cat on a neon rooftop
image: https://...

> /exit

Why three languages

Same protocol, same key, same endpoint — only the SDK differs. Pick whichever fits your stack.