main.py

import uvicorn
from fastapi import FastAPI, Request
from langchain_community.llms.ctransformers import CTransformers



MODEL_TYPE = 'mistral'
MODEL_BIN_PATH = "mistral-7b-instruct-v0.1.Q3_K_S.gguf"
MAX_NEW_TOKEN =  600
TEMPRATURE =  0.01
CONTEXT_LENGTH =  6000


llm = CTransformers(
            model=MODEL_BIN_PATH,
            config={
                'max_new_tokens': MAX_NEW_TOKEN,
                'temperature': TEMPRATURE,
                'context_length': CONTEXT_LENGTH
            },
            model_type=MODEL_TYPE
        )

app = FastAPI()


@app.post("//llm_on_cpu")
async def generate_text(request: Request):
    raw_data = await request.body()  # Get the raw body data from the request
    prompt = raw_data.decode('utf-8')
    if not prompt:
        return {'error': 'Prompt is required'}, 400
    try:
        generated_text = llm(prompt)
        print(generated_text)
        return {'generated_text': generated_text}
    except:
        return {"generated_text": "Error!"}