diff --git a/rag_tutorials/rag_database_routing/README.md b/rag_tutorials/rag_database_routing/README.md index 106c4b2..279f8a6 100644 --- a/rag_tutorials/rag_database_routing/README.md +++ b/rag_tutorials/rag_database_routing/README.md @@ -1,6 +1,6 @@ -# RAG Database Router Demo +# RAG Agent with Database Routing -This demo showcases RAG (Retrieval Augmented Generation) with database routing capabilities. The application allows users to: +This project showcases the RAG with database routing capabilities - which is a very efficient way to retrieve information from a large set of documents. The application allows users to: 1. Upload documents to three different databases: - Product Information @@ -9,6 +9,48 @@ This demo showcases RAG (Retrieval Augmented Generation) with database routing c 2. Query information using natural language, with automatic routing to the most relevant database. -## Setup +## Features -1. Create a virtual environment: +- **Document Upload**: Users can upload multiple PDF documents related to a particular company. These documents are processed and stored in one of the three databases: Product Information, Customer Support & FAQ, or Financial Information. + +- **Natural Language Querying**: Users can ask questions in natural language. The system automatically routes the query to the most relevant database using a phidata agent as the router. + +- **RAG Orchestration**: Utilizes Langchain for orchestrating the retrieval augmented generation process, ensuring that the most relevant information is retrieved and presented to the user. + +- **Fallback Mechanism**: If no relevant documents are found in the databases, a LangGraph agent with a DuckDuckGo search tool is used to perform web research and provide an answer. + +- **User Interface**: Built with Streamlit, providing an intuitive and interactive user experience. + +## How to Run? + +1. **Clone the Repository**: + ```bash + git clone https://github.com/Shubhamsaboo/awesome-llm-apps.git + cd rag_tutorials/rag_database_routing + ``` + +2. **Install Dependencies**: + ```bash + pip install -r requirements.txt + ``` + +3. **Run the Application**: + ```bash + streamlit run rag_database_routing.py + ``` + +4. **Configure API Key**: Obtain an OpenAI API key and set it in the application. This is required for initializing the language models used in the application. + +5. **Upload Documents**: Use the document upload section to add PDF documents to the desired database. + +6. **Ask Questions**: Enter your questions in the query section. The application will route your question to the appropriate database and provide an answer. + +## Technologies Used + +- **Langchain**: For RAG orchestration, ensuring efficient retrieval and generation of information. +- **Phidata Agent**: Used as the router agent to determine the most relevant database for a given query. +- **LangGraph Agent**: Acts as a fallback mechanism, utilizing DuckDuckGo for web research when necessary. +- **Streamlit**: Provides a user-friendly interface for document upload and querying. +- **ChromaDB**: Used for managing the databases, storing and retrieving document embeddings efficiently. + +This application is designed to streamline the process of retrieving information from large sets of documents, making it easier for users to find the answers they need quickly and efficiently. diff --git a/rag_tutorials/rag_database_routing/rag_database_routing.py b/rag_tutorials/rag_database_routing/rag_database_routing.py index fcd7723..6582225 100644 --- a/rag_tutorials/rag_database_routing/rag_database_routing.py +++ b/rag_tutorials/rag_database_routing/rag_database_routing.py @@ -1,25 +1,24 @@ import os -import getpass -from typing import List, Dict, Any, Literal +from typing import List, Dict, Literal from dataclasses import dataclass import streamlit as st -from dotenv import load_dotenv from langchain_core.documents import Document from langchain.text_splitter import RecursiveCharacterTextSplitter from langchain_community.document_loaders import PyPDFLoader from langchain_community.vectorstores import Chroma -from langchain_openai import OpenAIEmbeddings -from langchain.chains import LLMChain -from langchain_core.prompts import PromptTemplate -from langchain_openai import ChatOpenAI +from langchain_openai import OpenAIEmbeddings, ChatOpenAI import tempfile -from langchain_core.runnables import RunnableSequence -from langchain_core.output_parsers import StrOutputParser -from langchain_core.prompts import ChatPromptTemplate -from langchain_chroma import Chroma +from phi.agent import Agent +from phi.model.openai import OpenAIChat +from langchain.schema import HumanMessage +from langchain.chains.combine_documents import create_stuff_documents_chain +from langchain.chains import create_retrieval_chain +from langchain import hub +from langgraph.prebuilt import create_react_agent +from langchain_community.tools import DuckDuckGoSearchRun +from langchain_core.language_models import BaseLanguageModel def init_session_state(): - """Initialize session state variables""" if 'openai_api_key' not in st.session_state: st.session_state.openai_api_key = "" if 'embeddings' not in st.session_state: @@ -29,30 +28,11 @@ def init_session_state(): if 'databases' not in st.session_state: st.session_state.databases = {} -# Initialize session state at the top init_session_state() -# Constants -DatabaseType = Literal["products", "customer_support", "financials"] +DatabaseType = Literal["products", "support", "finance"] PERSIST_DIRECTORY = "db_storage" -ROUTER_TEMPLATE = """You are a query routing expert. Your job is to analyze user questions and determine which databases might contain relevant information. - -Available databases: -1. Product Information: Contains product details, specifications, and features -2. Customer Support & FAQ: Contains customer support information, frequently asked questions, and guides -3. Financial Information: Contains financial data, revenue, costs, and liabilities - -User question: {question} - -Return a comma-separated list of relevant databases (no spaces after commas). Only use these exact strings: -- products -- customer_support -- financials - -For example: "products,customer_support" if the question relates to both product info and support. -Your response:""" - @dataclass class CollectionConfig: name: str @@ -60,7 +40,6 @@ class CollectionConfig: collection_name: str persist_directory: str -# Collection configurations COLLECTIONS: Dict[DatabaseType, CollectionConfig] = { "products": CollectionConfig( name="Product Information", @@ -68,13 +47,13 @@ COLLECTIONS: Dict[DatabaseType, CollectionConfig] = { collection_name="products_collection", persist_directory=f"{PERSIST_DIRECTORY}/products" ), - "customer_support": CollectionConfig( + "support": CollectionConfig( name="Customer Support & FAQ", description="Customer support information, frequently asked questions, and guides", collection_name="support_collection", persist_directory=f"{PERSIST_DIRECTORY}/support" ), - "financials": CollectionConfig( + "finance": CollectionConfig( name="Financial Information", description="Financial data, revenue, costs, and liabilities", collection_name="finance_collection", @@ -83,47 +62,25 @@ COLLECTIONS: Dict[DatabaseType, CollectionConfig] = { } def initialize_models(): - """Initialize OpenAI models with API key""" if st.session_state.openai_api_key: - try: - os.environ["OPENAI_API_KEY"] = st.session_state.openai_api_key - # Test the API key with a small embedding request - test_embeddings = OpenAIEmbeddings(model="text-embedding-3-large") - test_embeddings.embed_query("test") - - # If successful, initialize the models - st.session_state.embeddings = test_embeddings - st.session_state.llm = ChatOpenAI(temperature=0) - st.session_state.databases = { - "products": Chroma( - collection_name=COLLECTIONS["products"].collection_name, - embedding_function=st.session_state.embeddings, - persist_directory=COLLECTIONS["products"].persist_directory - ), - "customer_support": Chroma( - collection_name=COLLECTIONS["customer_support"].collection_name, - embedding_function=st.session_state.embeddings, - persist_directory=COLLECTIONS["customer_support"].persist_directory - ), - "financials": Chroma( - collection_name=COLLECTIONS["financials"].collection_name, - embedding_function=st.session_state.embeddings, - persist_directory=COLLECTIONS["financials"].persist_directory - ) - } - return True - except Exception as e: - st.error(f"Error connecting to OpenAI API: {str(e)}") - st.error("Please check your internet connection and API key.") - return False + os.environ["OPENAI_API_KEY"] = st.session_state.openai_api_key + st.session_state.embeddings = OpenAIEmbeddings(model="text-embedding-3-large") + st.session_state.llm = ChatOpenAI(temperature=0) + + for config in COLLECTIONS.values(): + os.makedirs(config.persist_directory, exist_ok=True) + + st.session_state.databases = { + db_type: Chroma( + collection_name=config.collection_name, + embedding_function=st.session_state.embeddings, + persist_directory=config.persist_directory + ) for db_type, config in COLLECTIONS.items() + } + return True return False def process_document(file) -> List[Document]: - """Process uploaded PDF document""" - if not st.session_state.embeddings: - st.error("OpenAI API connection not initialized. Please check your API key.") - return [] - try: with tempfile.NamedTemporaryFile(delete=False, suffix='.pdf') as tmp_file: tmp_file.write(file.getvalue()) @@ -131,97 +88,99 @@ def process_document(file) -> List[Document]: loader = PyPDFLoader(tmp_path) documents = loader.load() - - # Clean up temporary file os.unlink(tmp_path) - text_splitter = RecursiveCharacterTextSplitter( - chunk_size=1000, - chunk_overlap=200 - ) - texts = text_splitter.split_documents(documents) - - return texts + text_splitter = RecursiveCharacterTextSplitter(chunk_size=800, chunk_overlap=300) + return text_splitter.split_documents(documents) except Exception as e: st.error(f"Error processing document: {e}") return [] -def route_query(question: str) -> List[DatabaseType]: - """Route the question to appropriate databases""" - router_prompt = ChatPromptTemplate.from_template(ROUTER_TEMPLATE) - router_chain = router_prompt | st.session_state.llm | StrOutputParser() - response = router_chain.invoke({"question": question}) - return response.strip().lower().split(",") - -def query_multiple_databases(question: str) -> str: - """Query multiple relevant databases and combine results""" - database_types = route_query(question) - all_docs = [] - - # Collect relevant documents from each database - for db_type in database_types: - db = st.session_state.databases[db_type] - docs = db.similarity_search(question, k=2) # Reduced k since we're querying multiple DBs - all_docs.extend(docs) - - # Sort all documents by relevance score if available - # Note: You might need to modify this based on your similarity search implementation - context = "\n\n---\n\n".join([doc.page_content for doc in all_docs]) - - answer_prompt = ChatPromptTemplate.from_template( - """Answer the question based on the following context from multiple databases. - If you use information from multiple sources, please indicate which type of source it came from. - If you cannot answer the question based on the context, say "I don't have enough information to answer this question." - -Context: {context} - -Question: {question} - -Answer:""" +def create_routing_agent() -> Agent: + return Agent( + model=OpenAIChat(id="gpt-4o", api_key=st.session_state.openai_api_key), + tools=[], + description="You are a query routing expert. Your only job is to analyze questions and determine which database they should be routed to.", + instructions=[ + "1. For questions about products, return 'products'", + "2. For questions about support, return 'support'", + "3. For questions about finance, return 'finance'", + "4. Return ONLY the database name" + ], + markdown=False, + show_tool_calls=False ) - - answer_chain = answer_prompt | st.session_state.llm | StrOutputParser() - return answer_chain.invoke({"context": context, "question": question}) -def clear_collection(collection_type: DatabaseType = None): - """Clear specified collection or all collections if none specified""" +def route_query(question: str) -> DatabaseType: try: - if collection_type: - if collection_type in st.session_state.databases: - collection_config = COLLECTIONS[collection_type] - # Delete collection - st.session_state.databases[collection_type]._collection.delete() - # Remove from session state - del st.session_state.databases[collection_type] - # Clean up persist directory - if os.path.exists(collection_config.persist_directory): - import shutil - shutil.rmtree(collection_config.persist_directory) - st.success(f"Cleared {collection_config.name} collection") - else: - # Clear all collections - for collection_type, collection_config in COLLECTIONS.items(): - if collection_type in st.session_state.databases: - st.session_state.databases[collection_type]._collection.delete() - if os.path.exists(collection_config.persist_directory): - import shutil - shutil.rmtree(collection_config.persist_directory) - st.session_state.databases = {} - st.success("Cleared all collections") + routing_agent = create_routing_agent() + response = routing_agent.run(question) + db_type = response.content.strip().lower().translate(str.maketrans('', '', '`\'"')) + + if db_type not in COLLECTIONS: + st.warning(f"Invalid database type: {db_type}, defaulting to products") + return "products" + + st.info(f"Routing question to {db_type} database") + return db_type except Exception as e: - st.error(f"Error clearing collection(s): {str(e)}") + st.error(f"Routing error: {str(e)}") + return "products" + +def create_fallback_agent(chat_model: BaseLanguageModel): + def web_research(query: str) -> str: + try: + search = DuckDuckGoSearchRun(num_results=5) + return search.run(query) + except Exception as e: + return f"Search failed: {str(e)}. Providing answer based on general knowledge." + + tools = [web_research] + return create_react_agent(model=chat_model, tools=tools, debug=False) + +def query_database(db: Chroma, question: str) -> tuple[str, list]: + try: + retriever = db.as_retriever(search_type="similarity_score_threshold", search_kwargs={"k": 4, "score_threshold": 0.4}) + relevant_docs = retriever.get_relevant_documents(question) + + if relevant_docs: + retrieval_qa_prompt = hub.pull("langchain-ai/retrieval-qa-chat") + combine_docs_chain = create_stuff_documents_chain(st.session_state.llm, retrieval_qa_prompt) + retrieval_chain = create_retrieval_chain(retriever, combine_docs_chain) + response = retrieval_chain.invoke({"input": question}) + return response['answer'], relevant_docs + + return _handle_web_fallback(question) + except Exception as e: + st.error(f"Error: {str(e)}") + return "I encountered an error. Please try rephrasing your question.", [] + +def _handle_web_fallback(question: str) -> tuple[str, list]: + st.info("No relevant documents found. Searching web...") + fallback_agent = create_fallback_agent(st.session_state.llm) + + with st.spinner('Researching...'): + agent_input = { + "messages": [HumanMessage(content=f"Research and provide a detailed answer for: '{question}'")], + "is_last_step": False + } + + try: + response = fallback_agent.invoke(agent_input, config={"recursion_limit": 100}) + if isinstance(response, dict) and "messages" in response: + answer = response["messages"][-1].content + return f"Web Search Result:\n{answer}", [] + except Exception: + fallback_response = st.session_state.llm.invoke(question).content + return f"Web search unavailable. General response: {fallback_response}", [] def main(): - st.title("📚 RAG with Database Routing") - + st.set_page_config(page_title="RAG Agent with Database Routing", page_icon="📚") + st.title("📚 RAG Agent with Database Routing") + with st.sidebar: st.header("Configuration") - api_key = st.text_input( - "Enter OpenAI API Key:", - type="password", - value=st.session_state.openai_api_key, - key="api_key_input" - ) + api_key = st.text_input("Enter OpenAI API Key:", type="password", value=st.session_state.openai_api_key, key="api_key_input") if api_key: st.session_state.openai_api_key = api_key @@ -233,53 +192,40 @@ def main(): if not st.session_state.openai_api_key: st.warning("Please enter your OpenAI API key to continue") st.stop() - - st.divider() - st.header("Database Management") - if st.button("Clear All Databases"): - clear_collection() - - st.divider() - st.subheader("Clear Individual Databases") - for collection_type, collection_config in COLLECTIONS.items(): - if st.button(f"Clear {collection_config.name}"): - clear_collection(collection_type) + + st.markdown("---") - # Document upload section st.header("Document Upload") - tabs = st.tabs([collection_config.name for collection_config in COLLECTIONS.values()]) + st.info("Upload documents to populate the databases. Each tab corresponds to a different database.") + tabs = st.tabs([config.name for config in COLLECTIONS.values()]) - for (collection_type, collection_config), tab in zip(COLLECTIONS.items(), tabs): + for (collection_type, config), tab in zip(COLLECTIONS.items(), tabs): with tab: - st.write(collection_config.description) - uploaded_file = st.file_uploader( - "Upload PDF document", - type="pdf", - key=f"upload_{collection_type}" - ) + st.write(config.description) + uploaded_files = st.file_uploader(f"Upload PDF documents to {config.name}", type="pdf", key=f"upload_{collection_type}", accept_multiple_files=True) - if uploaded_file: - with st.spinner('Processing document...'): - texts = process_document(uploaded_file) - if texts: + if uploaded_files: + with st.spinner('Processing documents...'): + all_texts = [] + for uploaded_file in uploaded_files: + texts = process_document(uploaded_file) + all_texts.extend(texts) + + if all_texts: db = st.session_state.databases[collection_type] - db.add_documents(texts) - st.success("Document processed and added to the database!") + db.add_documents(all_texts) + st.success("Documents processed and added to the database!") - # Query section st.header("Ask Questions") + st.info("Enter your question below to find answers from the relevant database.") question = st.text_input("Enter your question:") if question: with st.spinner('Finding answer...'): - # Get relevant databases - database_types = route_query(question) - - # Display routing information - st.info(f"Searching in: {', '.join([COLLECTIONS[db_type].name for db_type in database_types])}") - - # Get and display answer - answer = query_multiple_databases(question) + collection_type = route_query(question) + db = st.session_state.databases[collection_type] + st.info(f"Routing question to: {COLLECTIONS[collection_type].name}") + answer, relevant_docs = query_database(db, question) st.write("### Answer") st.write(answer) diff --git a/rag_tutorials/rag_database_routing/requirements.txt b/rag_tutorials/rag_database_routing/requirements.txt index 0ce6d76..c0259ec 100644 --- a/rag_tutorials/rag_database_routing/requirements.txt +++ b/rag_tutorials/rag_database_routing/requirements.txt @@ -1,9 +1,11 @@ -langchain>=0.1.0 -langchain-community>=0.0.10 -langchain-core>=0.1.10 -chromadb>=0.4.22 +langchain==0.3.12 +langchain-community==0.3.12 +langchain-core==0.3.28 +chromadb==0.5.20 streamlit>=1.29.0 -python-dotenv>=1.0.0 pypdf>=4.0.0 sentence-transformers>=2.2.2 -openai>=1.6.1 +phidata==2.7.3 +langchain-openai==0.2.14 +langgraph==0.2.53 +duckduckgo-search==6.4.1 \ No newline at end of file