app.py

from fastapi import FastAPI
from pydantic import BaseModel
from transformers import AutoModelForCausalLM, AutoTokenizer
from datasets import load_dataset
import torch

# Load model and tokenizer
MODEL_NAME = "meta-llama/Llama-3.2-1B"  # Replace with your fine-tuned model

tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, torch_dtype=torch.float16)

# Load AWS-Bot dataset
DATASET_NAME = "Faizal2805/cyberbot"  # Replace with your dataset
dataset = load_dataset(DATASET_NAME, split="train")

def get_dataset_response(prompt: str):
    """
    Check if the user's input matches a dataset entry and return a predefined response.
    If no match is found, return None.
    """
    for example in dataset:
        if example["text"].startswith(f"<s>[INST] {prompt} [/INST]"):
            return example["text"].split("</s>")[-1].strip()
    return None

def generate_response(prompt: str):
    """
    Generate a response from the dataset if available; otherwise, use the model.
    """
    dataset_response = get_dataset_response(prompt)
    if dataset_response:
        return dataset_response  # Return predefined dataset response

    # Fallback to model-based response
    inputs = tokenizer(prompt, return_tensors="pt")
    outputs = model.generate(**inputs, max_length=200)
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

app = FastAPI()

class Query(BaseModel):
    text: str

@app.post("/chat")
def chat(query: Query):
    response = generate_response(query.text)
    return {"response": response}

@app.get("/")
def root():
    return {"message": "AWS-Bot is running!"}