app.py

from datasets import load_dataset
import numpy as np
import gradio as gr
import chromadb

from transformers import AutoModel, AutoTokenizer, pipeline
import torch
import chromadb
import os
import requests

API_URL = "https://api-inference.huggingface.co/models/meta-llama/Llama-2-7b-hf"
headers = {"Authorization": f"Bearer {os.getenv('HF_Token')}"}

def query_llama(prompt):
    payload = {"inputs": prompt}
    response = requests.post(API_URL, headers=headers, json=payload)
    return response.json()

prompt = "Explain machine learning in simple terms."
response = query_llama(prompt)
print(response)


'''
# Initialize ChromaDB client
chroma_client = chromadb.PersistentClient(path="./chroma_db")  # Stores data persistently
collection = chroma_client.get_or_create_collection(name="wikipedia_docs")

# Load the BAAI embedding model
model_name = "BAAI/bge-base-en"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModel.from_pretrained(model_name)

def get_embedding(text):
    """Generate embeddings using BAAI/bge-base-en."""
    inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True)
    with torch.no_grad():
        outputs = model(**inputs)
    return outputs.last_hidden_state[:, 0, :].numpy().tolist()  # Take CLS token embedding

# Load LLaMA Model (Meta LLaMA 2)
#llama_pipe = pipeline("text-generation", model=AutoModelForCausalLM.from_pretrained("meta-llama/Llama-2-7b-chat-hf"))

# Load a small subset (10,000 rows)
#dataset = load_dataset("wiki40b", "en", split="train[:1000]")

# Extract only text
#docs = [d["text"] for d in dataset]
docs = ["Machine learning is a field of AI...", "Neural networks are inspired by the brain..."]

#print("Loaded dataset with", len(docs), "documents.")

# ✅ Step 2: Embed and Store in ChromaDB
for i, doc in enumerate(docs):
    embedding = get_embedding(doc)
    collection.add(ids=[str(i)], embeddings=[embedding], documents=[doc])

print("Stored embeddings in ChromaDB!")

# Store embeddings in ChromaDB
#for i, (doc, embedding) in enumerate(zip(docs, embeddings)):
#    collection.add(
#        ids=[str(i)],  # Unique ID for each doc
#        embeddings=[embedding.tolist()],  # Convert numpy array to list
#        documents=[doc]
#    )


# Search function using ChromaDB
#def search_wikipedia(query, top_k=3):
#    query_embedding = embed_model.encode([query]).tolist()
#    results = collection.query(
#        query_embeddings=query_embedding, 
#        n_results=top_k
    
    #return "\n\n".join(results["documents"][0])  # Return top results
#    return results["documents"][0]  # Return top results

# Function to search ChromaDB & generate response
def query_llama(user_input):
    query_embedding = get_embedding(user_input)
    results = collection.query(query_embeddings=[query_embedding], n_results=3)

    if not results["documents"]:
        return "No relevant documents found."

    context = " ".join(results["documents"][0])
    prompt = f"Using this context, answer the question: {user_input}\nContext: {context}"

    response = llama_pipe(prompt, max_length=200)
    return f"**LLaMA Response:** {response[0]['generated_text']}\n\n**Retrieved Docs:** {context}"


# Gradio Interface
iface = gr.Interface(
    fn=query_llama, 
    inputs="text", 
    outputs="text", 
    title="Wikipedia Search RAG",
    description="Enter a query and retrieve relevant Wikipedia passages."
)

iface.launch()
'''