Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -2,7 +2,7 @@ import os
|
|
| 2 |
import streamlit as st
|
| 3 |
import PyPDF2
|
| 4 |
import requests
|
| 5 |
-
|
| 6 |
import faiss
|
| 7 |
from groq import Groq
|
| 8 |
|
|
@@ -36,21 +36,13 @@ def chunk_text(text, max_length=500):
|
|
| 36 |
chunks.append(chunk.strip())
|
| 37 |
return chunks
|
| 38 |
|
| 39 |
-
# Function to compute embeddings using NumPy
|
| 40 |
-
def compute_embeddings(chunks):
|
| 41 |
-
embeddings = []
|
| 42 |
-
for chunk in chunks:
|
| 43 |
-
vector = np.array([ord(char) for char in chunk[:300]], dtype=np.float32) # Truncate to 300 characters
|
| 44 |
-
padded_vector = np.pad(vector, (0, 300 - len(vector)), mode="constant")
|
| 45 |
-
embeddings.append(padded_vector)
|
| 46 |
-
return np.vstack(embeddings)
|
| 47 |
-
|
| 48 |
# Function to create FAISS index
|
| 49 |
-
def create_faiss_index(
|
| 50 |
-
|
|
|
|
| 51 |
index = faiss.IndexFlatL2(dimension)
|
| 52 |
index.add(embeddings)
|
| 53 |
-
return index
|
| 54 |
|
| 55 |
# Function to query Groq API
|
| 56 |
def query_groq(question, model_name="llama-3.3-70b-versatile"):
|
|
@@ -70,17 +62,17 @@ def main():
|
|
| 70 |
st.write("Document text extracted successfully!")
|
| 71 |
|
| 72 |
st.write("Chunking and embedding text...")
|
|
|
|
| 73 |
chunks = chunk_text(text)
|
| 74 |
-
|
| 75 |
-
index = create_faiss_index(embeddings)
|
| 76 |
st.write(f"Created FAISS index with {len(chunks)} chunks.")
|
| 77 |
|
| 78 |
# Query input
|
| 79 |
question = st.text_input("Ask a question based on the document:")
|
| 80 |
if question:
|
| 81 |
st.write("Searching for relevant chunks...")
|
| 82 |
-
question_embedding =
|
| 83 |
-
|
| 84 |
relevant_chunk = chunks[indices[0][0]]
|
| 85 |
|
| 86 |
st.write("Generating answer using Groq API...")
|
|
@@ -92,3 +84,4 @@ if __name__ == "__main__":
|
|
| 92 |
main()
|
| 93 |
|
| 94 |
|
|
|
|
|
|
| 2 |
import streamlit as st
|
| 3 |
import PyPDF2
|
| 4 |
import requests
|
| 5 |
+
from sentence_transformers import SentenceTransformer
|
| 6 |
import faiss
|
| 7 |
from groq import Groq
|
| 8 |
|
|
|
|
| 36 |
chunks.append(chunk.strip())
|
| 37 |
return chunks
|
| 38 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
# Function to create FAISS index
|
| 40 |
+
def create_faiss_index(chunks, model):
|
| 41 |
+
embeddings = model.encode(chunks)
|
| 42 |
+
dimension = len(embeddings[0])
|
| 43 |
index = faiss.IndexFlatL2(dimension)
|
| 44 |
index.add(embeddings)
|
| 45 |
+
return index, chunks
|
| 46 |
|
| 47 |
# Function to query Groq API
|
| 48 |
def query_groq(question, model_name="llama-3.3-70b-versatile"):
|
|
|
|
| 62 |
st.write("Document text extracted successfully!")
|
| 63 |
|
| 64 |
st.write("Chunking and embedding text...")
|
| 65 |
+
model = SentenceTransformer("all-MiniLM-L6-v2")
|
| 66 |
chunks = chunk_text(text)
|
| 67 |
+
index, chunks = create_faiss_index(chunks, model)
|
|
|
|
| 68 |
st.write(f"Created FAISS index with {len(chunks)} chunks.")
|
| 69 |
|
| 70 |
# Query input
|
| 71 |
question = st.text_input("Ask a question based on the document:")
|
| 72 |
if question:
|
| 73 |
st.write("Searching for relevant chunks...")
|
| 74 |
+
question_embedding = model.encode([question])
|
| 75 |
+
_, indices = index.search(question_embedding, k=1)
|
| 76 |
relevant_chunk = chunks[indices[0][0]]
|
| 77 |
|
| 78 |
st.write("Generating answer using Groq API...")
|
|
|
|
| 84 |
main()
|
| 85 |
|
| 86 |
|
| 87 |
+
|