app.py

import gradio as gr
from sentence_transformers import SentenceTransformer
from sklearn.metrics.pairwise import cosine_similarity
import numpy as np
from transformers import pipeline

# doc set
docs = [
    "a python function is a reusable block of code. define one using 'def'. like: def greet(): print('hi')",
    "a for loop repeats code for each item in a list. like: for pet in pets: print(pet)",
    "a while loop repeats code while a condition is true. like: while x < 5: print(x); x += 1",
    "a variable stores a value. example: name = 'mei' stores the string 'mei' in the variable called name",
    "python is a programming language that's great for beginners. it’s used for everything from web apps to ai",
    "if statements let your code make choices. if it's raining: bring an umbrella! else: enjoy the sun",
    "import lets you pull in extra tools. like: import random lets you use random stuff in your code"
]

# embed docs
embedder = SentenceTransformer("all-MiniLM-L6-v2")
doc_vectors = embedder.encode(docs)

# use a Q&A pipeline instead of gen
qa = pipeline("question-answering")

def real_bot(message, history):
    # find closest doc
    q_vector = embedder.encode([message])
    sims = cosine_similarity(q_vector, doc_vectors)[0]
    best_idx = np.argmax(sims)
    context = docs[best_idx]

    # use qa model to answer
    result = qa(question=message, context=context)
    answer = result["answer"]
    
    return answer

# UI
chat = gr.ChatInterface(
    fn = real_bot,
    title = "mei's python bot 🐍",
    description = "ask me python related questions, i’ll keep it short & sweet!"
)

chat.launch()