changes

app.py

@@ -5,7 +5,7 @@ import os
 # Constants
 ROLE_USER = "user"
 ROLE_ASSISTANT = "assistant"
-MAX_MESSAGES = 4
+MAX_MESSAGES = 5
 
 st.set_page_config(page_title="Chat with Git", page_icon="🦜")
 st.title("Chat with Git 🤖📚")

src/main.py

@@ -13,7 +13,7 @@ from langchain.chat_models import ChatOpenAI
 from langchain.document_loaders import TextLoader
 from langchain.document_loaders import GitLoader
 from langchain.llms import OpenAI
-from langchain.memory import ConversationBufferMemory
+from langchain.memory import ConversationBufferMemory, ConversationBufferWindowMemory
 from langchain.vectorstores import Chroma
 from langchain.embeddings.openai import OpenAIEmbeddings
 from langchain.prompts import PromptTemplate, SystemMessagePromptTemplate, HumanMessagePromptTemplate, AIMessagePromptTemplate, ChatPromptTemplate
@@ -81,12 +81,17 @@ def retreival(vector_store, k):
     llm = ChatOpenAI(model=llm_name, temperature=0)
 
     # Define the system message template
+    #Adding CHAT HISTORY to the System template explicitly because mainly Chat history goes to Condense the Human Question with Backround (Not template), but System template goes straight the LLM Chain
+    #Explicitly adding chat history to access previous chats and answer "what is my previous question?"
+    #Great thing this also sends the chat history to the LLM Model along with the context and question
     system_template = """You're a code summarisation assistant. Given the following extracted parts of a long document as "CONTEXT" create a final answer.
     If you don't know the answer, just say that you don't know. Don't try to make up an answer.
     Only If asked to create a "DIAGRAM" for code use "MERMAID SYNTAX LANGUAGE" in your answer from "CONTEXT" and "CHAT HISTORY" with a short explanation of diagram.
 
     CONTEXT: {context}
     =======
+    CHAT HISTORY: {chat_history}
+    =======
     FINAL ANSWER:"""
 
     human_template = """{question}"""
@@ -104,11 +109,18 @@ def retreival(vector_store, k):
     PROMPT = ChatPromptTemplate.from_messages(messages)
 
     #Creating memory
-    memory = ConversationBufferMemory(
+    # memory = ConversationBufferMemory(
+    #         memory_key="chat_history",
+    #         input_key="question",
+    #         output_key="answer",
+    #         return_messages=True)
+
+    memory = ConversationBufferWindowMemory(
             memory_key="chat_history",
             input_key="question",
             output_key="answer",
-            return_messages=True)
+            return_messages=True,
+            k=5)
 
     #Creating the retriever, this can also be a contextual compressed retriever
     retriever = vector_store.as_retriever(search_type="similarity", search_kwargs={"k": k}) #search_type can be "similarity" or "mmr"
@@ -134,13 +146,9 @@ class ConversationalResponse:
         self.chunks = split_data(self.data)
         self.vector_store = ingest_chunks(self.chunks)
         self.chain_type = "stuff"
-        self.k = 15
+        self.k = 10
         self.chain = retreival(self.vector_store, self.k)
 
     def __call__(self, question):
-        chat_history = []
-        agent = self.chain({"question": question
-                            , "chat_history": chat_history
-                            })
-        chat_history.append((question, agent['answer']))
+        agent = self.chain(question)
         return agent['answer']