fastap, streamlit samples added

.gitignore

@@ -4,5 +4,7 @@
 .local
 .npm
 .cache
+.env
 shared
-data/model
+data/model
+__pycache__

README.md

@@ -36,7 +36,7 @@ After running the containers, you should be able to access to:
 * InterSystems IRIS [Management Portal](http://localhost:52774/csp/sys/UtilHome.csp). You can login using `superuser` / `SYS`
 * [Jupyter Notebook](http://localhost:8888) 
 
-# RAG Application
+# Explore RAG applications using Jupyter
 
 ## Medicine Leaflet examples
 
@@ -59,4 +59,54 @@ This example is about a company called Holefoods that sells food with some hole
 Using the sales data model of the company, the goal is to create an assistant that can translate natural language questions into valid SQL that answer the question.
 
 In [Jupyter Notebook](http://localhost:8888), you will find:
-* [QA-SQL-LLM.ipynb](./jupyter/QA-SQL-LLM.ipynb) - text to SQL example using OpenAI LLM.
+* [QA-SQL-LLM.ipynb](./jupyter/QA-SQL-LLM.ipynb) - text to SQL example using OpenAI LLM.
+
+# Create other applications
+
+There are some other examples you can try to build and modify in your local environment.
+
+First of all, create a new environment and install some requirements:
+
+```bash
+# create a local venv environment
+python3 -m venv .venv
+
+# activate venv
+source .venv/bin/activate
+
+# install dependencies
+pip3 install -r requirements.txt
+```
+
+Create an `.env` file for storing API keys for OpenAI / MistralAI. They will be used in the applications.
+
+```
+OPENAI_API_KEY="your-api"
+MISTRAL_API_KEY="your-api"
+```
+
+## Text to SQL service API 
+You can find a sample Text to SQL based on [QA-SQL-LLM.ipynb](./jupyter/QA-SQL-LLM.ipynb) [here](python/holefoods_text2sql/main.py).
+
+You can run it like this:
+
+```bash
+cd python/holefoods_text2sql
+fastapi dev main.py
+```
+
+Then open http://127.0.0.1:8000/docs to explore the API and try it out using the web client.
+
+## Streamlit Assistant
+There is also a great example of a langchain / streamlit chatbot assitant in https://alejandro-ao.com/how-to-use-streaming-in-langchain-and-streamlit/
+
+You can play with it here as well:
+
+```bash
+cd python/assitant
+streamlit run chatbot.py
+```
+
+Then open http://localhost:8501 and have a look at it.
+
+Are you able to add the logic to reproduce the Medicine Leaflet example in the assitant ?

docker-compose.yml

@@ -2,7 +2,6 @@ services:
 
   # iris
   iris:
-    init: true
     container_name: iris
     build:
       context: .
@@ -14,7 +13,6 @@ services:
     - ./shared:/shared
     environment:
     - ISC_DATA_DIRECTORY=/shared/durable
-    command: --check-caps false --ISCAgent false
 
   # jupyter notebook
   jupyter:

jupyter/requirements.txt

@@ -6,5 +6,6 @@ langchain-community
 mistral_common
 langchain_mistralai
 langchain-openai
+langchain_iris
 transformers
 torch

python/assistant/chatbot.py

@@ -0,0 +1,67 @@
+import streamlit as st
+from langchain_core.messages import AIMessage, HumanMessage
+#from langchain_openai import ChatOpenAI
+from langchain_mistralai import ChatMistralAI
+from dotenv import load_dotenv
+from langchain_core.output_parsers import StrOutputParser
+from langchain_core.prompts import ChatPromptTemplate
+
+load_dotenv()
+
+#llm = ChatOpenAI()
+llm = ChatMistralAI(model="mistral-large-latest")
+
+
+# app config
+st.set_page_config(page_title="Streamlit Chatbot", page_icon="🤖")
+st.title("Chatbot")
+
+
+def get_response(user_query, chat_history):
+
+    template = """
+    You are a helpful assistant. Answer the following questions considering the history of the conversation:
+
+    Chat history: {chat_history}
+
+    User question: {user_question}
+    """
+
+    prompt = ChatPromptTemplate.from_template(template)
+
+    chain = prompt | llm | StrOutputParser()
+
+    return chain.invoke({
+        "chat_history": chat_history,
+        "user_question": user_query,
+    })
+
+# session state
+if "chat_history" not in st.session_state:
+    st.session_state.chat_history = [
+        AIMessage(content="Hello, I am a bot. How can I help you?"),
+    ]
+
+
+# conversation
+for message in st.session_state.chat_history:
+    if isinstance(message, AIMessage):
+        with st.chat_message("AI"):
+            st.write(message.content)
+    elif isinstance(message, HumanMessage):
+        with st.chat_message("Human"):
+            st.write(message.content)
+
+# user input
+user_query = st.chat_input("Type your message here...")
+if user_query is not None and user_query != "":
+    st.session_state.chat_history.append(HumanMessage(content=user_query))
+
+    with st.chat_message("Human"):
+        st.markdown(user_query)
+
+    with st.chat_message("AI"):
+        response = get_response(user_query, st.session_state.chat_history)
+        st.write(response)
+
+    st.session_state.chat_history.append(AIMessage(content=response))

python/holefoods_text2sql/main.py

@@ -0,0 +1,124 @@
+from fastapi import FastAPI, Query
+from typing import Dict
+
+app = FastAPI(
+    title="Sample Holefoods schema Text to SQL API",
+    description="This is an example of a Text to SQL API using OpenAI LLM and IRIS Vector Database",
+    version="1.0.0"
+)
+
+# load OpenAI APIKEY from env
+from dotenv import load_dotenv
+load_dotenv()
+
+# database connection to extract information
+from langchain_community.utilities import SQLDatabase
+db = SQLDatabase.from_uri("iris://superuser:SYS@localhost:51774/LLMRAG", sample_rows_in_table_info=3, schema='Holefoods')
+
+# openai model
+from langchain_openai import ChatOpenAI
+llm = ChatOpenAI(model="gpt-4o-mini")
+
+from langchain_core.prompts import PromptTemplate
+
+# define the custom prompt template
+template = '''
+You are an InterSystems IRIS SQL expert. 
+Given an input question, first create a syntactically correct InterSystems IRIS SQL query to run and return the answer to the input question.
+Unless the user specifies in the question a specific number of examples to obtain, query for at most {top_k} results using the TOP as defined in InterSystems IRIS syntax: ```SELECT [DISTINCT] TOP int select-item, select-item,...```
+Always specify table names using schema as prefix.
+Do not use LIMIT clause as it is not correct in IRIS dialect.
+Do not end SQL sentences with an ;
+Do not enclose fields in quotes or double quotes.
+Do not enclose table names in quotes or double quotes.
+You can order the results to return the most informative data in the database.
+Never query for all columns from a table. You must query only the columns that are needed to answer the question.
+Pay attention to use only the column names you can see in the tables below. Be careful to not query for columns that do not exist. Also, pay attention to which column is in which table.
+Pay attention to use CAST(CURRENT_DATE as date) function to get the current date, if the question involves "today".
+
+Return only plain SQL without any formatting.
+
+Only use the following tables:
+
+{table_info}.
+Question: {input}'''
+
+# provide some examples to improve the results
+examples = [
+    { 
+        "input": "List all regions.", 
+        "query": "SELECT ID, Name FROM HoleFoods.Region;"
+    },
+    {
+        "input": "List all countries.",
+        "query": "SELECT c.ID, c.Name, r.Name Region FROM HoleFoods.Country c JOIN HoleFoods.Region r on c.Region=r.ID"
+    },
+    {
+        "input": "What are the different product categories ?",
+        "query": "SELECT DISTINCT(Category) Categories FROM HoleFoods.Product"
+    },
+    {
+        "input": "How many pasta products where sold online in 2023 ?",
+        "query": "SELECT SUM(UnitsSold) FROM HoleFoods.SalesTransaction st JOIN HoleFoods.Product p ON st.Product=p.ID WHERE st.Channel='Online' AND YEAR(st.DateOfSale) = 2023 AND p.Category = 'Pasta'"
+    },
+    {
+        "input": "Find all snack products",
+        "query": "SELECT SKU, Name, Price FROM HoleFoods.Product p WHERE p.Category='Snack'"
+    },
+    {
+        "input": "Find all candy products",
+        "query": "SELECT SKU, Name, Price FROM HoleFoods.Product p WHERE p.Category='Candy'"
+    },
+    {
+        "input": "How many products were sold in Europe in 2022 ?",
+        "query": "SELECT SUM(UnitsSold) FROM HoleFoods.SalesTransaction st JOIN HoleFoods.Outlet o ON st.Outlet=o.ID JOIN HoleFoods.Country c ON o.Country=c.ID JOIN HoleFoods.Region r ON c.Region=r.ID WHERE r.Name='Europe' AND YEAR(st.DateOfSale) = 2022"
+    }
+]
+
+from langchain_iris import IRISVector
+from langchain_core.example_selectors import SemanticSimilarityExampleSelector
+from langchain_openai import OpenAIEmbeddings
+
+example_selector = SemanticSimilarityExampleSelector.from_examples(
+    examples,
+    OpenAIEmbeddings(),
+    IRISVector,
+    k=3,
+    input_keys=["input"],
+    connection_string='iris://superuser:SYS@localhost:51774/LLMRAG',
+    collection_name="sql_samples",
+    pre_delete_collection=True
+)
+
+from langchain_core.prompts import FewShotPromptTemplate
+
+example_prompt = PromptTemplate.from_template("User input: {input}\nSQL query: {query}")
+
+prompt = FewShotPromptTemplate(
+    example_selector=example_selector,
+    example_prompt=example_prompt,
+    prefix=template,
+    suffix="User input: {input}\nSQL query: ",
+    input_variables=["input", "top_k", "table_info"],
+)
+
+from langchain.chains import create_sql_query_chain
+chain = create_sql_query_chain(llm, db, prompt)
+
+from pydantic import BaseModel, Field
+class Text2SQLResponse(BaseModel):
+    sql: str  # the generated SQL query
+
+@app.get("/text2sql")
+async def text2sql(text: str = Query(..., description="Natural language text about Holefoods schema to convert to SQL")) -> Text2SQLResponse:
+    """
+    Convert input text to a SQL query.
+
+    Parameters:
+    - text: The input natural language string.
+
+    Returns:
+    - JSON object containing the generated SQL query.
+    """
+    query = chain.invoke({"question": text})
+    return {"sql": query}

python/requirements.txt

@@ -0,0 +1,8 @@
+langchain-community
+langchain-openai
+langchain_iris
+fastapi[standard]
+streamlit
+mistral_common
+langchain_mistralai
+../jupyter/intersystems_irispython-3.2.0-py3-none-any.whl