addded summaries to metadata

config.py

@@ -1,15 +1,24 @@
 import os
 from dotenv import load_dotenv
 from rag_app.database.db_handler import DataBaseHandler
+from langchain_huggingface import HuggingFaceEndpoint
 
 load_dotenv()
 
 SQLITE_FILE_NAME = os.getenv('SOURCES_CACHE')
 PERSIST_DIRECTORY = os.getenv('VECTOR_DATABASE_LOCATION')
 EMBEDDING_MODEL = os.getenv("EMBEDDING_MODEL")
+SECONDARY_LLM_MODEL = os.getenv("SECONDARY_LLM_MODEL")
 
 
 db = DataBaseHandler()
 
 db.create_all_tables()
 
+SECONDARY_LLM = HuggingFaceEndpoint(
+        repo_id=SECONDARY_LLM_MODEL, 
+        temperature=0.1,         # Controls randomness in response generation (lower value means less random)
+        max_new_tokens=1024,     # Maximum number of new tokens to generate in responses
+        repetition_penalty=1.2,  # Penalty for repeating the same words (higher value increases penalty)
+        return_full_text=False   # If False, only the newly generated text is returned; if True, the input is included as well
+    )

rag_app/chains/__init__.py

@@ -1 +1,2 @@
-from rag_app.chains.user_response_sentiment_chain import user_response_sentiment_prompt
+from rag_app.chains.user_response_sentiment_chain import user_response_sentiment_prompt
+from rag_app.chains.generate_document_summary import generate_document_summary_prompt

rag_app/chains/generate_document_summary.py

@@ -0,0 +1,14 @@
+from langchain_core.prompts import PromptTemplate
+
+
+generate_document_summary_template = """ 
+You will be given a document object
+=================
+{document} 
+====================
+You must generate a summary 
+
+
+"""
+
+generate_document_summary_prompt = PromptTemplate.from_template(generate_document_summary_template)

rag_app/knowledge_base/utils.py

@@ -1,7 +1,39 @@
 from langchain_core.documents import Document
+from chains import generate_document_summary_prompt
+from config import SECONDARY_LLM
 
 
 def generate_document_summaries(
         docs: list[Document]
-    ):
-    pass
+    ) -> list[Document]:
+    """
+    Generates summaries for a list of Document objects and updates their metadata with the summaries.
+
+    Args:
+        docs (List[Document]): A list of Document objects to generate summaries for.
+
+    Returns:
+        List[Document]: A new list of Document objects with updated metadata containing the summaries.
+
+    Example:
+        docs = [Document(metadata={"title": "Doc1"}), Document(metadata={"title": "Doc2"})]
+        updated_docs = generate_document_summaries(docs)
+        for doc in updated_docs:
+            print(doc.metadata["summary"])
+
+    """
+    
+    new_docs = docs.copy()
+    
+    for doc in new_docs:
+        
+        genrate_summary_chain = generate_document_summary_prompt | SECONDARY_LLM
+        summary = genrate_summary_chain.invoke(
+            {"document":str(doc.metadata)}
+        )        
+        
+        doc.metadata.update(
+            {"summary":summary}
+        )
+    
+    return new_docs