qdrant feature added

app.py

@@ -2,13 +2,17 @@ import streamlit as st
 from langchain import memory as lc_memory
 from langsmith import Client
 from streamlit_feedback import streamlit_feedback
-from utils import get_expression_chain, retriever
+from utils import get_expression_chain, retriever, get_embeddings, create_qdrant_collection
 from langchain_core.tracers.context import collect_runs
+from qdrant_client import QdrantClient
 from dotenv import load_dotenv
+import os
 
 load_dotenv()
 client = Client()
-
+qdrant_api=os.getenv("QDRANT_API_KEY")
+qdrant_url=os.getenv("QDRANT_URL")
+qdrant_client = QdrantClient(qdrant_url ,api_key=qdrant_api)
 st.set_page_config(page_title = "SUP'ASSISTANT")
 st.subheader("Hey there! How can I help you today!")
 
@@ -47,13 +51,33 @@ if prompt := st.chat_input(placeholder="What do you need to know about SUP'COM ?
         input_dict = {"input": prompt}
 
         with collect_runs() as cb:
-            for chunk in chain.stream(input_dict, config={"tags": ["Streamlit Chat"]}):
+            for chunk in chain.stream(input_dict, config={"tags": ["SUP'ASSISTANT"]}):
                 full_response += chunk.content
                 message_placeholder.markdown(full_response + "▌")
             memory.save_context(input_dict, {"output": full_response})
             st.session_state.run_id = cb.traced_runs[0].id
         message_placeholder.markdown(full_response)
 
+        run_id = st.session_state.run_id
+        question_embedding = get_embeddings(prompt)
+        answer_embedding = get_embeddings(full_response)
+        # Add question and answer to Qdrant
+        qdrant_client.upload_collection(            
+            collection_name="chat-history",
+            payload=[
+                {"text": prompt, "type": "question", "question_ID": run_id},
+                {"text": full_response, "type": "answer", "question_ID": run_id}
+            ],
+            vectors=[
+                question_embedding,
+                answer_embedding,
+            ],
+            parallel=4,
+            max_retries=3,
+            )
+
+        
+
 if st.session_state.get("run_id"):
     run_id = st.session_state.run_id
     feedback = streamlit_feedback(
@@ -93,4 +117,24 @@ if st.session_state.get("run_id"):
                 "score": score,
             }
         else:
-            st.warning("Invalid feedback score.")
+            st.warning("Invalid feedback score.")
+        if feedback.get("text"):
+            comment = feedback.get("text")
+            feedback_embedding = get_embeddings(comment)
+        else:
+            comment = "no comment"
+            feedback_embedding = get_embeddings(comment)
+
+        
+        qdrant_client.upload_collection(            
+            collection_name="chat-history",
+            payload=[
+                {"text": comment,"Score:":score, "type": "feedback", "question_ID": run_id}
+            ],
+            vectors=[
+                feedback_embedding
+            ],
+            parallel=4,
+            max_retries=3,
+            )
+            

requirements.txt

@@ -10,4 +10,5 @@ langchain-community
 rank_bm25
 cohere
 nomic[local]
-streamlit-feedback
+streamlit-feedback
+qdrant-client

utils.py

@@ -9,8 +9,10 @@ from dotenv import load_dotenv
 from langchain_core.prompts import ChatPromptTemplate
 from langchain_core.runnables import Runnable, RunnableMap
 from langchain.schema import BaseRetriever
-load_dotenv()
+from qdrant_client import models
 
+load_dotenv()
+#Retriever
 def retriever(n_docs=5):
     vector_database_path = "sup-knowledge-eng-nomic"
 
@@ -44,6 +46,7 @@ def retriever(n_docs=5):
 
     return retriever
 
+#Retriever prompt
 rag_prompt = """You are an assistant for question-answering tasks.
 The questions that you will be asked will mainly be about SUP'COM (also known as Higher School Of Communication Of Tunis).
 
@@ -65,6 +68,7 @@ Answer:"""
 def format_docs(docs):
     return "\n\n".join(doc.page_content for doc in docs)
 
+#RAG chain
 def get_expression_chain(retriever: BaseRetriever, model_name="llama-3.1-70b-versatile", temp=0
 ) -> Runnable:
     """Return a chain defined primarily in LangChain Expression Language"""
@@ -90,4 +94,18 @@ def get_expression_chain(retriever: BaseRetriever, model_name="llama-3.1-70b-ver
     llm = ChatGroq(model=model_name, temperature=temp)
 
     chain = ingress | prompt | llm
-    return chain
+    return chain
+
+embedding_model = NomicEmbeddings(model="nomic-embed-text-v1.5", inference_mode="local")
+#Generate embeddings for a given text
+def get_embeddings(text):
+    return embedding_model.embed([text], task_type='search_document')[0]
+
+
+# Create or connect to a Qdrant collection
+def create_qdrant_collection(client, collection_name):
+    if collection_name not in client.get_collections().collections:
+        client.create_collection(
+            collection_name=collection_name,
+            vectors_config=models.VectorParams(size=768, distance=models.Distance.COSINE)
+        )