UPDATE: ParentDocumentRetreiver

functions.py

@@ -3,6 +3,9 @@ from langchain_text_splitters import RecursiveCharacterTextSplitter
 from langchain_qdrant import QdrantVectorStore
 from langchain_core.prompts.chat import ChatPromptTemplate
 from langchain_core.output_parsers import StrOutputParser
+from langchain.retrievers import ParentDocumentRetriever
+from langchain.storage import InMemoryStore
+from langchain.docstore.document import Document
 from langchain_huggingface import HuggingFaceEmbeddings
 from supabase.client import create_client
 from qdrant_client import QdrantClient
@@ -40,6 +43,7 @@ Context:
 NOTE: generate responses WITHOUT prepending phrases like "Response:", "Output:", or "Answer:", etc
 """
 prompt = ChatPromptTemplate.from_template(prompt)
+store = InMemoryStore()
 
 
 def createUser(username: str, password: str) -> None:
@@ -100,19 +104,31 @@ def createTable(tablename: str):
 def addDocuments(text: str, vectorstore: str):
     try:
         global embeddings
-        text_splitter = RecursiveCharacterTextSplitter(
-            chunk_size = 1024,
-            chunk_overlap = 200,
+        global store
+        parentSplitter = RecursiveCharacterTextSplitter(
+            chunk_size = 2400,
+            chunk_overlap = 240,
             add_start_index = True
         )
-        texts = text_splitter.create_documents([text])
+        childSplitter = RecursiveCharacterTextSplitter(
+            chunk_size = 300,
+            chunk_overlap = 90,
+            add_start_index = True
+        )
+        texts = [Document(page_content = text)]
         vectorstore = QdrantVectorStore.from_existing_collection(
             embedding = embeddings,
             collection_name=vectorstore,
             url=os.environ["QDRANT_URL"],
             api_key=os.environ["QDRANT_API_KEY"]
         )
-        vectorstore.add_documents(documents = texts)
+        retriever = ParentDocumentRetriever(
+            vectorstore=vectorstore,
+            docstore=store,
+            child_splitter=childSplitter,
+            parent_splitter=parentSplitter
+        )
+        retriever.add_documents(documents = texts)
         return {
             "output": "SUCCESS"
         }
@@ -139,11 +155,15 @@ def answerQuery(query: str, vectorstore: str, llmModel: str = "llama3-70b-8192")
         url=os.environ["QDRANT_URL"],
         api_key=os.environ["QDRANT_API_KEY"]
     )
-    retriever = vectorstore.as_retriever()
+    retriever = ParentDocumentRetriever(
+        vectorstore=vectorstore,
+        docstore=store,
+        child_splitter=RecursiveCharacterTextSplitter()
+    )
     chain = (
         {"context": retriever | RunnableLambda(format_docs), "question": RunnablePassthrough()}
         | prompt
-        | ChatGroq(model = llmModel, temperature = 0.3, max_tokens = 512)
+        | ChatGroq(model = llmModel, temperature = 0.5, max_tokens = 512)
         | StrOutputParser()
         )
     return {
@@ -168,7 +188,7 @@ def listTables(username: str):
         global qdrantClient
         qdrantCollections = qdrantClient.get_collections()
         return {
-            "output": list(filter(lambda x: True if x.split("-")[1] == username else False, [x.name for x in qdrantCollections.collections]))
+            "output": list(filter(lambda x: True if x.split("-")[1] == username else False, [x.name.split("-")[-1] for x in qdrantCollections.collections]))
         }
     except Exception as e:
         return {