from transformers import pipeline
from langchain_cohere import ChatCohere
from langchain_core.messages import HumanMessage, SystemMessage 
from langchain_core.output_parsers import StrOutputParser
import gradio as gr

llm = ChatCohere(model='command-r')
pipe = pipeline("automatic-speech-recognition", model="openai/whisper-base")
parser = StrOutputParser()

def getting_prompt(txtmsg,spkmsg):
    recog_text = pipe(spkmsg)
    messages = [
        SystemMessage(content=txtmsg),
        HumanMessage(content=recog_text['text']),
    ]
    chain = llm | parser
    response = chain.invoke(messages)
    return response

demo = gr.Interface(getting_prompt,['text','audio'],'text')
demo.launch()