app.py

import streamlit as st
import pandas as pd
from torch import cuda
from langchain_community.embeddings import HuggingFaceEmbeddings, HuggingFaceInferenceAPIEmbeddings
from langchain_community.vectorstores import Qdrant
from qdrant_client import QdrantClient
from langchain.retrievers import ContextualCompressionRetriever
from langchain.retrievers.document_compressors import CrossEncoderReranker
from langchain_community.cross_encoders import HuggingFaceCrossEncoder
from appStore.prep_data import process_giz_worldwide
from appStore.prep_utils import create_documents
from appStore.embed import hybrid_embed_chunks, get_local_qdrant

# get the device to be used eithe gpu or cpu
device = 'cuda' if cuda.is_available() else 'cpu'


st.set_page_config(page_title="SEARCH IATI",layout='wide')
st.title("SEARCH IATI Database")
var=st.text_input("enter keyword")
  
####################  Create the embeddings collection and save ######################
# the steps below need to be performed only once and then commented out any unnecssary compute over-run
##### First we process and create the chunks for relvant data source
#chunks = process_giz_worldwide()
##### Convert to langchain documents
#temp_doc = create_documents(chunks,'chunks')
##### Embed and store docs, check if collection exist then you need to update the collection
#hybrid_embed_chunks(docs= temp_doc, collection_name = "giz_worldwide")

################### Hybrid Search ######################################################



button=st.button("search")
#found_docs = vectorstore.similarity_search(var)
#print(found_docs)
# results= get_context(vectorstore, f"find the relvant paragraphs for: {var}")
if button:
     st.write(f"Found {len(results)} results for query:{var}")

     for i in results: 
         st.subheader(str(i.metadata['id'])+":"+str(i.metadata['title_main']))
         st.caption(f"Status:{str(i.metadata['status'])}, Country:{str(i.metadata['country_name'])}")
         st.write(i.page_content)
         st.divider()