Update with simplified app

README.md

@@ -1,25 +1,22 @@
 ## Web Wanderer
 
-This is a lightweight app for the LangChain [Web Research Retriever](https://github.com/langchain-ai/langchain/pull/8102).
+This is a lightweight app using the [Web Research Retriever](https://github.com/langchain-ai/langchain/pull/8102).
 
 You only need to supply a few thiings.
 
+In `settings()` function, supply:
+
 ### Search
-Supply search functionality e.g., Google - 
-```
-export GOOGLE_CSE_ID=xxx
-export GOOGLE_API_KEY=xxx
-search = GoogleSearchAPIWrapper() 
-```
+Select the search tool you want to use (e.g., GoogleSearchAPIWrapper).
 
-### Public API
-Supply API key(s) e.g., OpenAI -
-```
-export OPENAI_API_KEY=sk-xxx
-```
+### Vectorstore
+Select the vectorstore you want to use (e.g., Chroma).
+
+### LLM
+Select the vectorstore you want to use (e.g., ChatOpenAI).
 
-### Private
-Follow [setup](https://python.langchain.com/docs/use_cases/question_answering/local_retrieval_qa) for local LLMs and supply path. 
+Then, run:
 
-### Run
-streamlit run web_wanderer.py
+```
+streamlit run web_explorer.py
+```

web_explorer.py

@@ -0,0 +1,59 @@
+import os
+import streamlit as st
+from langchain.chains import RetrievalQAWithSourcesChain
+from langchain.retrievers.web_research import WebResearchRetriever
+
+@st.cache_resource
+def settings():
+
+    # Vectorstore
+    from langchain.vectorstores import Chroma
+    from langchain.embeddings.openai import OpenAIEmbeddings
+    vectorstore_public = Chroma(embedding_function=OpenAIEmbeddings())
+
+    # LLM
+    from langchain.chat_models import ChatOpenAI
+    llm = ChatOpenAI(temperature=0)
+
+    # Search
+    from langchain.utilities import GoogleSearchAPIWrapper
+    os.environ["GOOGLE_CSE_ID"] = "xxx"
+    os.environ["GOOGLE_API_KEY"] = "xxx"
+    search = GoogleSearchAPIWrapper()   
+
+    # Initialize 
+    web_retriever = WebResearchRetriever.from_llm(
+    vectorstore=vectorstore_public,
+    llm=llm, 
+    search=search, 
+    )
+
+    return web_retriever, llm
+
+st.sidebar.image("img/ai.png")
+st.header("`Interweb Explorer`")
+st.info("`I am an AI that can answer questions by exploring, reading, and summarizing web pages."
+    "I can be configured to use different moddes: public API or private (no data sharing).`")
+
+# Make retriever and llm
+web_retriever, llm = settings()
+
+# User input 
+question = st.text_input("`Ask a question:`")
+
+if question:
+
+    # Generate answer (w/ citations)
+    import logging
+    logging.basicConfig()
+    logging.getLogger("langchain.retrievers.web_research").setLevel(logging.INFO)    
+    qa_chain = RetrievalQAWithSourcesChain.from_chain_type(llm,
+                                                           retriever=web_retriever)
+
+    # Write answer and sources
+    result = qa_chain({"question": question})
+    st.info('`Answer:`')
+    st.info(result['answer'])
+    st.info('`Source:`')
+    st.info(result['sources'])
+