Update app.py

app.py

@@ -1,90 +1,60 @@
-import streamlit as st
-import requests
-import os
-from sentence_transformers import SentenceTransformer
-import faiss
-import numpy as np
+import sys
+!{sys.executable} -m pip install streamlit transformers accelerate torch
 
-st.set_page_config(page_title="📚 AI Adaptive Learning (Ultimate)", layout="wide")
-st.title("📚 AI Adaptive Learning & Smart Revision System (Mistral 7B)")
+import streamlit as st
 
-# -----------------------------
-# 1️⃣ Hugging Face API Setup
-# -----------------------------
-HF_API_TOKEN = st.secrets["HF_API_TOKEN"]
+HF_API_TOKEN = "your_huggingface_api_token_here"  # <-- paste your token
 MODEL_ID = "mistralai/Mistral-7B-Instruct-v0.2"
 API_URL = f"https://api-inference.huggingface.co/models/{MODEL_ID}"
 HEADERS = {"Authorization": f"Bearer {HF_API_TOKEN}"}
 
-def query_hf_api(prompt):
-    payload = {
-        "inputs": prompt,
-        "parameters": {"max_new_tokens": 250, "temperature": 0.7, "top_p": 0.9}
-    }
-    response = requests.post(API_URL, headers=HEADERS, json=payload)
-    if response.status_code == 200:
-        output = response.json()
-        return output[0]["generated_text"]
-    else:
-        return f"Error: {response.status_code} - {response.text}"
-
-# -----------------------------
-# 2️⃣ Build RAG (Context Retrieval)
-# -----------------------------
-@st.cache_resource
-def build_rag_index(doc_folder="docs"):
-    embeddings = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
-    corpus = []
-    corpus_texts = []
-
-    for file_name in os.listdir(doc_folder):
-        if file_name.endswith(".txt"):
-            path = os.path.join(doc_folder, file_name)
-            with open(path, "r", encoding="utf-8") as f:
-                text = f.read()
-                sentences = text.split("\n")
-                for sent in sentences:
-                    if sent.strip():
-                        corpus.append(embeddings.encode(sent))
-                        corpus_texts.append(sent)
-
-    dim = corpus[0].shape[0]
-    index = faiss.IndexFlatL2(dim)
-    index.add(np.array(corpus).astype("float32"))
-    return index, corpus_texts, embeddings
-
-index, corpus_texts, embedder = build_rag_index()
-
-# -----------------------------
-# 3️⃣ Retrieve Context
-# -----------------------------
-def retrieve_context(query, k=3):
-    query_vec = embedder.encode(query).astype("float32")
-    D, I = index.search(np.array([query_vec]), k)
-    context = "\n".join([corpus_texts[i] for i in I[0]])
-    return context
-
-# -----------------------------
-# 4️⃣ Streamlit UI
-# -----------------------------
-difficulty = st.selectbox("Select explanation style:", ["Explain Like I'm 10", "Detailed", "Step-by-Step"])
-user_question = st.text_input("Ask a question:")
-
-if st.button("Submit") and user_question:
-    with st.spinner("Generating high-quality answer..."):
-        context = retrieve_context(user_question, k=3)
-        prompt = f"""
-You are a knowledgeable teacher and explain concepts clearly.
+import streamlit as st
+from transformers import AutoTokenizer, AutoModelForCausalLM, AutoConfig
+import torch
 
-Context:
-{context}
+st.title("📚 AI Adaptive Learning (Local Small Model)")
 
-Question:
-{user_question}
+MODEL_ID = "microsoft/phi-2"
 
-Explain the answer in a {difficulty} manner.
-Answer:
-"""
-        answer = query_hf_api(prompt)
-        st.subheader("📖 AI Answer:")
-        st.write(answer)
+@st.cache_resource
+def load_model():
+    tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
+    # Set pad_token_id for the tokenizer if it's not already set, using eos_token_id as a fallback
+    if tokenizer.pad_token_id is None:
+        tokenizer.pad_token_id = tokenizer.eos_token_id
+
+    # Load the model configuration
+    config = AutoConfig.from_pretrained(MODEL_ID)
+
+    # Check if pad_token_id exists in config and set it if not, for Phi models this often needs to be explicitly added
+    if not hasattr(config, 'pad_token_id') or config.pad_token_id is None:
+        config.pad_token_id = tokenizer.pad_token_id
+
+    model = AutoModelForCausalLM.from_pretrained(
+        MODEL_ID,
+        config=config, # Pass the modified config to the model
+        torch_dtype=torch.float32,
+        device_map="auto"
+    )
+    return tokenizer, model
+
+tokenizer, model = load_model()
+
+# Input question
+user_input = st.text_input("Ask a question:")
+
+if st.button("Submit") and user_input:
+    inputs = tokenizer(user_input, return_tensors="pt")
+
+    with torch.no_grad():
+        outputs = model.generate(
+            **inputs,
+            max_new_tokens=150,
+            do_sample=True,
+            temperature=0.7
+        )
+
+    answer = tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+    st.subheader("AI Answer:")
+    st.write(answer)