Update app.py

app.py

@@ -1,15 +1,14 @@
-from langchain_community.llms import Ollama
-from langchain_community.vectorstores import Chroma
-from langchain_community.embeddings import SentenceTransformerEmbeddings
+# Necessary imports
+from langchain.vectorstores import Chroma
+from langchain.embeddings import HuggingFaceEmbeddings
 from langchain.text_splitter import CharacterTextSplitter
 from langchain.prompts import PromptTemplate
 from langchain.chains.question_answering import load_qa_chain
 from datasets import load_dataset
 import pandas as pd
 from functools import lru_cache
-from langchain_huggingface import HuggingFaceEmbeddings
-import gradio as gr
 from huggingface_hub import InferenceClient
+import gradio as gr
 
 # Initialize the Hugging Face Inference Client
 client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
@@ -18,26 +17,34 @@ client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
 dataset = load_dataset('arbml/LK_Hadith')
 df = pd.DataFrame(dataset['train'])
 
-# Filter data
+# Filter data (Only retain Hadiths with non-weak grades)
 filtered_df = df[df['Arabic_Grade'] != 'ضعيف']
 documents = list(filtered_df['Arabic_Matn'])
 metadatas = [{"Hadith_Grade": grade} for grade in filtered_df['Arabic_Grade']]
 
-# Use CharacterTextSplitter
-text_splitter = CharacterTextSplitter(chunk_size=10000)
+# Text splitter (using a smaller chunk size for memory efficiency)
+text_splitter = CharacterTextSplitter(chunk_size=1000)
 nltk_chunks = text_splitter.create_documents(documents, metadatas=metadatas)
 
-# LLM
-llm = Ollama(model="salmatrafi/acegpt:7b")
+# LLM (Replace Ollama with a Hugging Face Hub model)
+from langchain.llms import HuggingFaceHub
+llm = HuggingFaceHub(repo_id="salmatrafi/acegpt:7b")
 
-# Create an embedding model
+# Create an embedding model (Hugging Face transformer model for embeddings)
 embeddings = HuggingFaceEmbeddings(model_name="intfloat/multilingual-e5-base")
 
+# Generate document embeddings
 docs_text = [doc.page_content for doc in nltk_chunks]
-docs_embedding = embeddings.embed_documents(docs_text)
+try:
+    docs_embedding = embeddings.embed_documents(docs_text)
+except Exception as e:
+    print(f"Error in embedding generation: {str(e)}")
 
-# Create Chroma vector store
-vector_store = Chroma.from_documents(nltk_chunks, embedding=embeddings)
+# Create Chroma vector store with embeddings
+try:
+    vector_store = Chroma.from_documents(nltk_chunks, embedding=embeddings)
+except Exception as e:
+    print(f"Error in creating vector store: {str(e)}")
 
 # Question answering prompt template
 qna_template = "\n".join([
@@ -134,16 +141,19 @@ def respond(
 
     response = ""
 
-    for msg in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = msg.choices[0].delta.content
-        response += token
-        yield response
+    try:
+        for msg in client.chat_completion(
+            messages,
+            max_tokens=max_tokens,
+            stream=True,
+            temperature=temperature,
+            top_p=top_p,
+        ):
+            token = msg.choices[0].delta.content
+            response += token
+            yield response
+    except Exception as e:
+        yield f"An error occurred during chat completion: {str(e)}"
 
 # Gradio Chat Interface
 demo = gr.ChatInterface(
@@ -162,5 +172,6 @@ demo = gr.ChatInterface(
     ],
 )
 
+# Launch the Gradio interface
 if __name__ == "__main__":
     demo.launch()