Readme LCEL update and rag.py docstrings

Author: rauni98iitr

assets/lcel_pipe_flow.png

@@ -13,10 +13,9 @@
 from langchain.callbacks.base import BaseCallbackHandler
 from langchain_core.runnables import RunnablePassthrough
 from langchain_core.output_parsers import StrOutputParser
-
 from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler
 
-# for streaming in Streamlit without LECL
+################### for streaming in Streamlit without LECL ###################
 class StreamHandler(BaseCallbackHandler):
     def __init__(self, container, initial_text=""):
         self.container = container
@@ -25,7 +24,12 @@ def __init__(self, container, initial_text=""):
     def on_llm_new_token(self, token: str, **kwargs) -> None:
         self.text += token
         self.container.markdown(self.text)
-
+# stream_handler = StreamHandler(st.empty())
+""" if you want to use streaming on your streamlit app, it's tricky to seperate model script \n
+and streamlit script if not using LECL, because llm will have to use 'streaming=True' \n
+and 'callbacks=[stream_handler]' and streamhandler uses st.empty() placeholder here which can't be first streamlit command. 
+"""
+               
 ####################### Data processing for vectorstore #################################
 pdf_folder_path = "./data_source"
 documents = []
@@ -91,7 +95,6 @@ def format_docs(docs):
 
 n_gpu_layers = 1
 n_batch = 512
-# stream_handler = StreamHandler(st.empty())
 
 llm = LlamaCpp(
     model_path="/Users/raunakanand/Documents/Work_R/llm_models/mistral-7b-v0.1.Q4_K_S.gguf",
@@ -105,7 +108,7 @@ def format_docs(docs):
     # callbacks=[StreamingStdOutCallbackHandler()]
 )
 
-########## When using RetrievalQA chain from llm's chain ##########
+########## use when using RetrievalQA chain from llm's chain ##########
 qa = RetrievalQA.from_chain_type(llm=llm, chain_type='stuff',
                                  retriever=retriever,
                                 #  return_source_documents=True,

assets/snap1.png

@@ -1,13 +1,22 @@
-pypdf
-langchain
-transformers
-chromadb
-streamlit
-sentence-transformers
-# Example: METAL
-CMAKE_ARGS="-DLLAMA_METAL=on"  FORCE_CMAKE=1 pip install llama-cpp-python==0.1.83 --no-cache-dir
+# About
 
+This project runs a local llm agent based RAG model on langchain using new pipesyntax [LCEL](https://python.langchain.com/docs/expression_language/get_started)(LangChain Expression Language) as well as older LLM chains(RetrievalQA), see `rag.py`. <br> We are using LECL in rag.py for inference as it has a smooth output streaming generator output which is consumed by streamlit using 'write_stream' method.
 
+The model uses persistent ChromaDB for vector store, which takes all the pdf files in `data_source` directory (one pdf about titanic for demo).
+
+The UI is built on streamlit, where the output of RAG model is streamed token on the streamlit app in a chat format, see `st_app.py`.
+
+![image info](./assets/snap1.png)
+
+### <u>LCEL - LangChain Expression Language</u>:
+Langchain composes chain of components in linux pip system like:</br>
+`chain = retriever | prompt | llm | Outputparser` </br>
+See implementation in `rag.py`
+
+![image info](./assets/lcel_pipe_flow.png)
+
+
+For more: [Pinecone LCEL Article](https://www.pinecone.io/learn/series/langchain/langchain-expression-language/)
 
 # Enviornment Setup 
 

rag.py

@@ -3,9 +3,7 @@
 
 st.set_page_config(page_title="LLM Search Titaninc", page_icon=':robot:')
 # st.header("Query PDF")
-st.title("Welcome")
-
-# prompt = st.chat_input("Enter your message...")
+st.title("Welcome to Langchain RAG")
 
 if ('messages' not in st.session_state):
     st.session_state.messages = []
@@ -14,7 +12,7 @@
     with st.chat_message(message["role"]):
         st.markdown(message["content"])
 
-prompt = st.chat_input("Enter your message...")
+prompt = st.chat_input("Enter your query about Titanic...")
 
 if (prompt):
     st.session_state.messages.append({'role':'user', 'content': prompt})