Use the Qdrant vector db instead of the json file

Author: zioproto

infra/aks.tf

@@ -38,9 +38,9 @@ module "aks" {
     acr = azurerm_container_registry.acr.id
   }
 
-  network_policy                 = "azure"
-  net_profile_dns_service_ip     = "10.0.0.10"
-  net_profile_service_cidr       = "10.0.0.0/16"
+  network_policy             = "azure"
+  net_profile_dns_service_ip = "10.0.0.10"
+  net_profile_service_cidr   = "10.0.0.0/16"
 
   key_vault_secrets_provider_enabled = true
   secret_rotation_enabled            = true

infra/identity.tf

@@ -1,4 +1,4 @@
-resource azurerm_user_assigned_identity chatbot {
+resource "azurerm_user_assigned_identity" "chatbot" {
   name                = "chatbot"
   resource_group_name = azurerm_resource_group.this.name
   location            = azurerm_resource_group.this.location

infra/installation_script.tftpl

@@ -1,5 +1,12 @@
 #!/bin/bash
 az aks get-credentials --resource-group ${resourceGroup} --name ${aksName} --overwrite-existing
+
+kubectl create namespace chatbot --dry-run=client -o yaml | kubectl apply -f -
+
+helm repo add qdrant https://qdrant.github.io/qdrant-helm
+helm repo update
+helm upgrade --install  --namespace chatbot --set image.tag=v1.11.3 chatbot qdrant/qdrant
+
 az acr build --registry ${registry} --image chatbot:latest ./../sample-application
 
 az identity federated-credential create --name chatbotfederatedidentity --identity-name chatbot --resource-group ${resourceGroup} --issuer ${oidc_url} --subject system:serviceaccount:chatbot:chatbot

infra/variables.tf

@@ -14,7 +14,7 @@ variable "chat_model_version" {
 }
 
 variable "scale_type" {
-  type    = string
+  type        = string
   description = "values: GlobalStandard, Standard"
-  default = "GlobalStandard"
+  default     = "GlobalStandard"
 }

sample-application/chatbot.py

@@ -20,17 +20,18 @@
 from langchain.chat_models import AzureChatOpenAI
 from langchain.embeddings import AzureOpenAIEmbeddings
 from azure.identity import ManagedIdentityCredential
+import qdrant_client
 
 import streamlit as st
+
+from llama_index.vector_stores.qdrant import QdrantVectorStore
 from llama_index.core import (
     SimpleDirectoryReader,
-    GPTVectorStoreIndex,
+    VectorStoreIndex,
     PromptHelper,
     ServiceContext,
     StorageContext,
-    load_index_from_storage,
     Settings
-
 )
 
 from llama_index.llms.langchain import LangChainLLM
@@ -39,13 +40,26 @@
 
 from dotenv import load_dotenv, dotenv_values
 
-
 logging.basicConfig(stream=sys.stdout, level=logging.INFO)
 logging.getLogger("llama_index").setLevel(logging.DEBUG)
 
 index = None
 doc_path = "./data/"
-index_file = "index.json"
+
+client = qdrant_client.QdrantClient(
+    # you can use :memory: mode for fast and light-weight experiments,
+    # it does not require to have Qdrant deployed anywhere
+    # but requires qdrant-client >= 1.1.1
+    # location=":memory:"
+    # otherwise set Qdrant instance address with:
+    # url="http://<host>:<port>"
+    # otherwise set Qdrant instance with host and port:
+    host="chatbot-qdrant",
+    port=6333
+    # set API KEY for Qdrant Cloud
+    # api_key="<qdrant-api-key>",
+)
+
 
 if "config" not in st.session_state:
     # Read the environment variables
@@ -119,47 +133,48 @@ def send_click():
 )
 
 if uploaded_file and uploaded_file.name != st.session_state.current_file:
-    # Ingest the document and create the index
-    with st.spinner('Ingesting the file..'):
-        doc_files = os.listdir(doc_path)
-        for doc_file in doc_files:
-            os.remove(doc_path + doc_file)
-
-        bytes_data = uploaded_file.read()
-        with open(f"{doc_path}{uploaded_file.name}", "wb") as f:
-            f.write(bytes_data)
-
-        loader = SimpleDirectoryReader(doc_path, recursive=True, exclude_hidden=True)
-        documents = loader.load_data()
-        sidebar_placeholder.header("Current Processing Document:")
-        sidebar_placeholder.subheader(uploaded_file.name)
-        sidebar_placeholder.write(documents[0].get_text()[:500] + "...")
-
-        index = GPTVectorStoreIndex.from_documents(
-            documents, service_context=service_context
-        )
-
-        index.set_index_id("vector_index")
-        index.storage_context.persist(index_file)
-        st.session_state.current_file = uploaded_file.name
-        st.session_state.response = ""  # clean up the response when new file is uploaded
-    st.success('Done!')
-
-elif os.path.exists(index_file):
-    # Read from storage context
-    storage_context = StorageContext.from_defaults(persist_dir=index_file)
-    index = load_index_from_storage(
-        storage_context, index_id="vector_index", service_context=service_context
+    st.session_state.current_file = uploaded_file.name
+    st.session_state.response = ""  # clean up the response when new file is uploaded
+    if not client.collection_exists(collection_name=uploaded_file.name):
+        # Ingest the document and create the index
+        with st.spinner('Ingesting the file..'):
+            doc_files = os.listdir(doc_path)
+            for doc_file in doc_files:
+                os.remove(doc_path + doc_file)
+
+            bytes_data = uploaded_file.read()
+            with open(f"{doc_path}{uploaded_file.name}", "wb") as f:
+                f.write(bytes_data)
+
+            loader = SimpleDirectoryReader(doc_path, recursive=True, exclude_hidden=True)
+            documents = loader.load_data()
+            sidebar_placeholder.header("Current Processing Document:")
+            sidebar_placeholder.subheader(uploaded_file.name)
+
+            vector_store = QdrantVectorStore(client=client, collection_name=uploaded_file.name)
+            storage_context = StorageContext.from_defaults(vector_store=vector_store)
+            index = VectorStoreIndex.from_documents(
+                documents,
+                service_context=service_context,
+                storage_context=storage_context,
+            )
+            index.set_index_id("vector_index")
+        st.success('Done!')
+
+if st.session_state.current_file:
+    vector_store = QdrantVectorStore(client=client, collection_name=uploaded_file.name)
+    storage_context = StorageContext.from_defaults(vector_store=vector_store)
+    index = VectorStoreIndex.from_vector_store(
+        vector_store=vector_store,
+        service_context=service_context,
+        storage_context=storage_context,
+        index_id="vector_index",
     )
-
-    loader = SimpleDirectoryReader(doc_path, recursive=True, exclude_hidden=True)
-    documents = loader.load_data()
-    doc_filename = os.listdir(doc_path)[0]
+    doc_filename = st.session_state.current_file
     sidebar_placeholder.header("Current Processing Document:")
-    sidebar_placeholder.subheader(doc_filename)
-    sidebar_placeholder.write(documents[0].get_text()[:500] + "...")
+    sidebar_placeholder.subheader(uploaded_file.name)
 
-if index:
+if index or st.session_state.response != "":
     st.text_input("Ask something: ", key="prompt", on_change=send_click)
     st.button("Send", on_click=send_click)
     if st.session_state.response:

sample-application/requirements.txt

@@ -9,3 +9,4 @@ python-dotenv==1.0.0
 pypdf==4.1.0
 azure-core==1.26.4
 azure-identity==1.13.0
+llama-index-vector-stores-qdrant==0.2.17