import os
import gradio as gr
import modelscope_gradio_components as mgr
from modelscope_gradio_components.components.Chatbot.llm_thinking_presets import \
qwen
def resolve_assets(relative_path):
return os.path.join(os.path.dirname(__file__), "../resources",
relative_path)
conversation = [
[
None, {
"text": f"""
标签语法:
```json
{{"text": "风和日丽", "resolution": "1024*1024"}}
```
qwen preset:
Action: image_gen
Action Input: {{"text": "风和日丽", "resolution": "1024*1024"}}
Observation: ![IMAGEGEN]({resolve_assets("screen.jpeg")}) 根据您的描述"风和日丽",我生成了一张图片。![]({resolve_assets("screen.jpeg")})
Action: 「任意文本表示,将展示为思考链调用的名称」
Action Input: 「任意json or md 内容,将展示到调用过程的下拉框」
Observation: 「任意 md 内容,将作为完成调用的展示的下拉框内」
""",
"flushing": False
}
],
]
with gr.Blocks() as demo:
mgr.Chatbot(
value=conversation,
llm_thinking_presets=[qwen()],
height=600,
)
if __name__ == "__main__":
demo.queue().launch()