Migrate to Llama Index 0.5.X (#9)

* Migrate to llama_index 0.5.X, update apps

* Update README with spaces links

Author: logan-markewich

README.md

@@ -40,14 +40,16 @@ There are two main example folders
   - loads text from the documents folder (using `st.cache_resource`, so it only loads once)
   - provides an input text-box and a button to run the query
   - the string response is displayed after it finishes
-  - want to see this example in action? Check it out [here](https://llama-index.streamlit.app/)
+  - want to see this example in action? Check it out [here](https://huggingface.co/spaces/llamaindex/llama_index_vector_demo)
 
 - streamlit_sql_sandbox (runs on localhost:8501)
   - `streamlit run streamlit_demo.py`
   - creates a streamlit app using a local SQL database about restaurant inspections in San Francisco ([data sample](https://docs.google.com/spreadsheets/d/1Ag5DBIviYiuRrt2yr3nXmbPFV-FOg5fDH5SM3ZEDnpw/edit#gid=780513932))
   - The "Setup" tab allows you to configure various LLM and LLama Index settings
   - The "Llama Index" tab demos some basic Text2SQL capabilities using only Llama Index
   - The "Langchain+Llama Index" tab uses a custom langchain agent, and uses the SQL index from Llama Index as a tool during conversations.
+  - Check out the huggingface space [here!](https://huggingface.co/spaces/llamaindex/llama_index_sql_sandbox)
+
 
 ## Docker
 Each example contains a `Dockerfile`. You can run `docker build -t my_tag_name .` to build a python3.11-slim docker image inside your desired folder. It ends up being about 600MB-900MB depending on the example.

flask_react/index_server.py

@@ -6,7 +6,7 @@
 
 from multiprocessing import Lock
 from multiprocessing.managers import BaseManager
-from llama_index import SimpleDirectoryReader, GPTSimpleVectorIndex, Document
+from llama_index import SimpleDirectoryReader, GPTSimpleVectorIndex, Document, ServiceContext
 
 index = None
 stored_docs = {}
@@ -19,12 +19,13 @@
 def initialize_index():
     """Create a new global index, or load one from the pre-set path."""
     global index, stored_docs
-
+    
+    service_context = ServiceContext.from_defaults(chunk_size_limit=512)
     with lock:
         if os.path.exists(index_name):
-            index = GPTSimpleVectorIndex.load_from_disk(index_name)
+            index = GPTSimpleVectorIndex.load_from_disk(index_name, service_context=service_context)
         else:
-            index = GPTSimpleVectorIndex([])
+            index = GPTSimpleVectorIndex([], service_context=service_context)
             index.save_to_disk(index_name)
         if os.path.exists(pkl_name):
             with open(pkl_name, "rb") as f:
@@ -44,11 +45,11 @@ def insert_into_index(doc_file_path, doc_id=None):
     document = SimpleDirectoryReader(input_files=[doc_file_path]).load_data()[0]
     if doc_id is not None:
         document.doc_id = doc_id
-    
-    # Keep track of stored docs -- llama_index doesn't make this easy
-    stored_docs[document.doc_id] = document.text[0:200]  # only take the first 200 chars
 
     with lock:
+        # Keep track of stored docs -- llama_index doesn't make this easy
+        stored_docs[document.doc_id] = document.text[0:200]  # only take the first 200 chars
+
         index.insert(document)
         index.save_to_disk(index_name)
 

flask_react/launch_app.sh

@@ -10,4 +10,5 @@ sleep 60
 # start the flask server
 python ./flask_demo.py &
 
+# assumes you've ran npm install already (dockerfile does this during build)
 cd react_frontend && npm run build && serve -s build

flask_react/requirements.txt

@@ -1,5 +1,5 @@
 Flask==2.2.3
 Flask-Cors==3.0.10
-langchain==0.0.123
-llama-index==0.4.39
+langchain==0.0.128
+llama-index==0.5.4
 PyPDF2==3.0.1

streamlit_sql_sandbox/requirements.txt

@@ -1,4 +1,4 @@
-langchain==0.0.123
-llama-index==0.4.39
+langchain==0.0.128
+llama-index==0.5.4
 streamlit==1.19.0
 streamlit-chat==0.0.2.2

streamlit_sql_sandbox/streamlit_demo.py

@@ -6,7 +6,7 @@
 from langchain.agents import Tool, initialize_agent
 from langchain.chains.conversation.memory import ConversationBufferMemory
 
-from llama_index import GPTSQLStructStoreIndex, LLMPredictor
+from llama_index import GPTSQLStructStoreIndex, LLMPredictor, ServiceContext
 from llama_index import SQLDatabase as llama_SQLDatabase
 from llama_index.indices.struct_store import SQLContextContainerBuilder
 
@@ -19,16 +19,11 @@
 )
 from utils import get_sql_index_tool, get_llm
 
-# NOTE: for local testing only, do NOT deploy with your key hardcoded
-# to use this for yourself, create a file called .streamlit/secrets.toml with your api key
-# Learn more about Streamlit on the docs: https://docs.streamlit.io/
-os.environ["OPENAI_API_KEY"] = st.secrets["openai_api_key"]
-
 
 @st.cache_resource
-def initialize_index(llm_name, model_temperature, table_context_dict, sql_path=DEFAULT_SQL_PATH):
+def initialize_index(llm_name, model_temperature, table_context_dict, api_key, sql_path=DEFAULT_SQL_PATH):
     """Create the GPTSQLStructStoreIndex object."""
-    llm = get_llm(llm_name, model_temperature)
+    llm = get_llm(llm_name, model_temperature, api_key)
 
     engine = create_engine(sql_path)
     sql_database = llama_SQLDatabase(engine)
@@ -37,23 +32,24 @@ def initialize_index(llm_name, model_temperature, table_context_dict, sql_path=D
     if table_context_dict is not None:
         context_builder = SQLContextContainerBuilder(sql_database, context_dict=table_context_dict)
         context_container = context_builder.build_context_container()
-
+    
+    service_context = ServiceContext.from_defaults(llm_predictor=LLMPredictor(llm=llm))
     index = GPTSQLStructStoreIndex([], 
                                    sql_database=sql_database, 
                                    sql_context_container=context_container, 
-                                   llm_predictor=LLMPredictor(llm=llm))
+                                   service_context=service_context)
 
     return index
 
 
 @st.cache_resource
-def initialize_chain(llm_name, model_temperature, lc_descrp, _sql_index):
+def initialize_chain(llm_name, model_temperature, lc_descrp, api_key, _sql_index):
     """Create a (rather hacky) custom agent and sql_index tool."""
     sql_tool = Tool(name="SQL Index", 
                     func=get_sql_index_tool(_sql_index, _sql_index.sql_context_container.context_dict), 
                     description=lc_descrp)
 
-    llm = get_llm(llm_name, model_temperature)
+    llm = get_llm(llm_name, model_temperature, api_key=api_key)
 
     memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)
 
@@ -64,19 +60,23 @@ def initialize_chain(llm_name, model_temperature, lc_descrp, _sql_index):
 
 st.title("🦙 Llama Index SQL Sandbox 🦙")
 st.markdown((
-    "This sandbox uses a sqlite database by default, containing information on health violations and inspections at restaurants in San Francisco.\n\n"
-    "The database contains three tables - businesses, inspections, and violations.\n\n"
+    "This sandbox uses a sqlite database by default, powered by [Llama Index](https://gpt-index.readthedocs.io/en/latest/index.html) ChatGPT, and LangChain.\n\n" 
+    "The database contains information on health violations and inspections at restaurants in San Francisco."
+    "This data is spread across three tables - businesses, inspections, and violations.\n\n"
     "Using the setup page, you can adjust LLM settings, change the context for the SQL tables, and change the tool description for Langchain."
+    "The other tabs will perform chatbot and text2sql operations.\n\n"
+    "Read more about LlamaIndexes structured data support [here!](https://gpt-index.readthedocs.io/en/latest/guides/tutorials/sql_guide.html)"
 ))
 
 
 setup_tab, llama_tab, lc_tab = st.tabs(["Setup", "Llama Index", "Langchain+Llama Index"])
 
 with setup_tab:
     st.subheader("LLM Setup")
-    model_temperature = st.slider("LLM Temperature", min_value=0.0, max_value=1.0, step=0.1)
+    api_key = st.text_input("Enter your OpenAI API key here", type="password")
     llm_name = st.selectbox('Which LLM?', ["text-davinci-003", "gpt-3.5-turbo", "gpt-4"])
-    
+    model_temperature = st.slider("LLM Temperature", min_value=0.0, max_value=1.0, step=0.1)
+
     st.subheader("Table Setup")
     business_table_descrp = st.text_area("Business table description", value=DEFAULT_BUSINESS_TABLE_DESCRP)
     violations_table_descrp = st.text_area("Business table description", value=DEFAULT_VIOLATIONS_TABLE_DESCRP)
@@ -92,10 +92,10 @@ def initialize_chain(llm_name, model_temperature, lc_descrp, _sql_index):
 with llama_tab:
     st.subheader("Text2SQL with Llama Index")
     if st.button("Initialize Index", key="init_index_1"):
-        st.session_state['llama_index'] = initialize_index(llm_name, model_temperature, table_context_dict if use_table_descrp else None)
+        st.session_state['llama_index'] = initialize_index(llm_name, model_temperature, table_context_dict if use_table_descrp else None, api_key)
 
     if "llama_index" in st.session_state:
-        query_text = st.text_input("Query:")
+        query_text = st.text_input("Query:", value="Which restaurant has the most violations?")
         if st.button("Run Query") and query_text:
             with st.spinner("Getting response..."):
                 try:
@@ -119,11 +119,11 @@ def initialize_chain(llm_name, model_temperature, lc_descrp, _sql_index):
     st.subheader("Langchain + Llama Index SQL Demo")
 
     if st.button("Initialize Agent"):
-        st.session_state['llama_index'] = initialize_index(llm_name, model_temperature, table_context_dict if use_table_descrp else None)
-        st.session_state['lc_agent'] = initialize_chain(llm_name, model_temperature, lc_descrp, st.session_state['llama_index'])
+        st.session_state['llama_index'] = initialize_index(llm_name, model_temperature, table_context_dict if use_table_descrp else None, api_key)
+        st.session_state['lc_agent'] = initialize_chain(llm_name, model_temperature, lc_descrp, api_key, st.session_state['llama_index'])
         st.session_state['chat_history'] = []
 
-    model_input = st.text_input("Message:")
+    model_input = st.text_input("Message:", value="Which restaurant has the most violations?")
     if 'lc_agent' in st.session_state and st.button("Send"):
         model_input = "User: " + model_input
         st.session_state['chat_history'].append(model_input)
@@ -134,4 +134,4 @@ def initialize_chain(llm_name, model_temperature, lc_descrp, _sql_index):
     if 'chat_history' in st.session_state:
         for msg in st.session_state['chat_history']:
             st_message(msg.split("User: ")[-1], is_user="User: " in msg)
-        
+        

streamlit_sql_sandbox/utils.py

@@ -1,3 +1,4 @@
+import os
 from langchain import OpenAI
 from langchain.chat_models import ChatOpenAI
 
@@ -16,9 +17,9 @@ def run_sql_index_query(query_text):
     return run_sql_index_query
 
 
-
-def get_llm(llm_name, model_temperature):
+def get_llm(llm_name, model_temperature, api_key):
+    os.environ['OPENAI_API_KEY'] = api_key
     if llm_name == "text-davinci-003":
         return OpenAI(temperature=model_temperature, model_name=llm_name)
     else:
-        return ChatOpenAI(temperature=model_temperature, model_name=llm_name)
+        return ChatOpenAI(temperature=model_temperature, model_name=llm_name)

streamlit_vector/requirements.txt

@@ -1,3 +1,3 @@
-langchain==0.0.123
-llama-index==0.4.39
+langchain==0.0.128
+llama-index==0.5.4
 streamlit==1.19.0

streamlit_vector/streamlit_demo.py

@@ -1,12 +1,7 @@
 import os
-
 import streamlit as st
-from llama_index import GPTSimpleVectorIndex, SimpleDirectoryReader
-
-# NOTE: for local testing only, do NOT deploy with your key hardcoded
-# to use this for yourself, create a file called .streamlit/secrets.toml with your api key
-# Learn more about Streamlit on the docs: https://docs.streamlit.io/
-os.environ["OPENAI_API_KEY"] = st.secrets["openai_api_key"]
+from llama_index import GPTSimpleVectorIndex, SimpleDirectoryReader, ServiceContext
+from llama_index.llm_predictor.chatgpt import ChatGPTLLMPredictor
 
 
 index_name = "./index.json"
@@ -15,11 +10,13 @@
 
 @st.cache_resource
 def initialize_index(index_name, documents_folder):
+    llm_predictor = ChatGPTLLMPredictor()
+    service_context = ServiceContext.from_defaults(llm_predictor=llm_predictor)
     if os.path.exists(index_name):
-        index = GPTSimpleVectorIndex.load_from_disk(index_name)
+        index = GPTSimpleVectorIndex.load_from_disk(index_name, service_context=service_context)
     else:
         documents = SimpleDirectoryReader(documents_folder).load_data()
-        index = GPTSimpleVectorIndex(documents)
+        index = GPTSimpleVectorIndex.from_documents(documents, service_context=service_context)
         index.save_to_disk(index_name)
 
     return index
@@ -31,16 +28,29 @@ def query_index(_index, query_text):
     return str(response)
 
 
-# This should be cached and only fully runs once
-index = initialize_index(index_name, documents_folder)
-
-
 st.title("🦙 Llama Index Demo 🦙")
 st.header("Welcome to the Llama Index Streamlit Demo")
-st.text("Please enter a query about Paul Graham's essays")
+st.write("Enter a query about Paul Graham's essays. You can check out the original essay [here](https://raw.githubusercontent.com/jerryjliu/llama_index/main/examples/paul_graham_essay/data/paul_graham_essay.txt). Your query will be answered using the essay as context, using embeddings from text-ada-002 and LLM completions from ChatGPT. You can read more about Llama Index and how this works in [our docs!](https://gpt-index.readthedocs.io/en/latest/index.html)")
+
+index = None
+api_key = st.text_input("Enter your OpenAI API key here:", type="password")
+if api_key:
+    os.environ['OPENAI_API_KEY'] = api_key
+    index = initialize_index(index_name, documents_folder)    
+
+
+if index is None:
+    st.warning("Please enter your api key first.")
 
-text = st.text_input("Query text:")
+text = st.text_input("Query text:", value="What did the author do growing up?")
 
 if st.button("Run Query") and text is not None:
     response = query_index(index, text)
     st.markdown(response)
+    
+    llm_col, embed_col = st.columns(2)
+    with llm_col:
+        st.markdown(f"LLM Tokens Used: {index.service_context.llm_predictor._last_token_usage}")
+    
+    with embed_col:
+        st.markdown(f"Embedding Tokens Used: {index.service_context.embed_model._last_token_usage}")