rag to streamlit + new pdf

Modules/rag.py

@@ -9,17 +9,24 @@ from langchain_community.document_loaders import PyPDFLoader
 from langchain.text_splitter import RecursiveCharacterTextSplitter
 from langchain_community.document_loaders import WebBaseLoader
 from langchain_community.vectorstores import Chroma, FAISS
+from langchain.chains.combine_documents import create_stuff_documents_chain
 from langchain_mistralai import MistralAIEmbeddings
 from langchain import hub
+from langchain.chains import (
+    create_history_aware_retriever,
+    create_retrieval_chain,
+)
 from typing import Literal
-from langchain_core.prompts import ChatPromptTemplate
+from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
 from langchain_core.pydantic_v1 import BaseModel, Field
 from langchain_mistralai import ChatMistralAI
 from langchain.embeddings.sentence_transformer import SentenceTransformerEmbeddings
 from langchain_community.tools import DuckDuckGoSearchRun
+from pathlib import Path
 
 def load_chunk_persist_pdf() -> Chroma:
-    pdf_folder_path = "data/pdf_folder/"
+    
+    pdf_folder_path = os.path.join(os.getcwd(),Path("data/pdf/"))
     documents = []
     for file in os.listdir(pdf_folder_path):
         if file.endswith('.pdf'):
@@ -32,7 +39,7 @@ def load_chunk_persist_pdf() -> Chroma:
     vectorstore = Chroma.from_documents(
         documents=chunked_documents,
         embedding=MistralAIEmbeddings(),
-        persist_directory="data/chroma_store/"
+        persist_directory= os.path.join(os.getcwd(),Path("data/chroma_store/"))
     )
     vectorstore.persist()
     return vectorstore
@@ -54,26 +61,29 @@ class RouteQuery(BaseModel):
 # LLM with function call 
 llm = ChatMistralAI(model="mistral-large-latest", mistral_api_key=mistral_api_key, temperature=0)
 
-# structured_llm_router = llm.with_structured_output(RouteQuery, method="json_mode")
-
-# Prompt 
-system = """You are an expert at routing a user question to a vectorstore or web search.
-The vectorstore contains documents related to agents, prompt engineering, and adversarial attacks.
-Use the vectorstore for questions on these topics. For all else, use web-search."""
-route_prompt = ChatPromptTemplate.from_messages(
-    [
-        ("system", system),
-        ("human", "{question}"),
-    ]
+
+prompt = ChatPromptTemplate.from_template(
+    """
+    You are a professional AI coach specialized in fitness, bodybuilding and nutrition.
+    You must adapt to the user : if he is a beginner, use simple words. You are gentle and motivative.
+    Use the following pieces of retrieved context to answer the question.
+    If you don't know the answer, just say that you don't know, and to refer to a nutritionist or a doctor.
+    Use three sentences maximum and keep the answer concise.
+
+    Question: {question} 
+
+    Context: {context} 
+
+    Answer:
+    """, 
 )
-prompt = hub.pull("rlm/rag-prompt")
 from langchain_core.output_parsers import StrOutputParser
 from langchain_core.runnables import RunnablePassthrough
 
 def format_docs(docs):
         return "\n\n".join(doc.page_content for doc in docs)
     
-    
+
 rag_chain = (
         {"context": retriever | format_docs, "question": RunnablePassthrough()}
         | prompt
@@ -81,6 +91,7 @@ rag_chain = (
         | StrOutputParser()
     )
 
-print(rag_chain.invoke("Build a fitness program for me. Be precise in terms of exercises"))
+
+# print(rag_chain.invoke("Build a fitness program for me. Be precise in terms of exercises"))
 
 # print(rag_chain.invoke("I am a 45 years old woman and I have to loose weight for the summer. Provide me with a fitness program"))

app.py

@@ -6,8 +6,19 @@ from langchain_mistralai import ChatMistralAI
 from dotenv import load_dotenv
 load_dotenv() # load .env api keys 
 import os
+
+from Modules.rag import rag_chain
+
 mistral_api_key = os.getenv("MISTRAL_API_KEY")
 
+def format_messages(messages):
+    formatted_messages = ""
+    for message in messages:
+        role = message["role"]
+        content = message["content"]
+        formatted_messages += f"{role}: {content}\n"
+    return formatted_messages
+
 st.set_page_config(layout="wide", initial_sidebar_state="collapsed")
 # Create two columns
 col1, col2 = st.columns(2)
@@ -43,8 +54,10 @@ with col1:
 
         with st.chat_message("assistant"):
             # Build answer from LLM
-            
-            response = llm.invoke(st.session_state.messages).content
+            response = rag_chain.invoke(
+                        instruction
+                        )
+            print(type(response))
             st.session_state.messages.append({"role": "assistant", "content": response})
             st.markdown(response)
 

data/pdf/F12_Strength&Conditioning_Program.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b6d7c1c04d0a98433e00e4a3ce1586311164a3ac50fc0e14a8fffb65ca7356b
+size 17579128