main.py

import os
import dotenv
import gradio as gr

# LLMs
import openai
from langchain.chat_models import AzureChatOpenAI
from langchain.llms import AzureOpenAI
from langchain.schema import AIMessage, HumanMessage
from langchain_experimental.sql import SQLDatabaseChain
from langchain.agents import create_sql_agent
from langchain.agents.agent_toolkits import SQLDatabaseToolkit
from langchain.agents.agent_types import AgentType
from langchain.agents.mrkl.prompt import FORMAT_INSTRUCTIONS

# Databases
import adal
import struct
from sqlalchemy.engine import URL
from sqlalchemy import create_engine, event
from langchain import SQLDatabase

# Load environment variables
dotenv.load_dotenv()

# Set OpenAI API settings from the .env file
openai.api_type = "azure"
openai.api_base = os.getenv("OPENAI_API_BASE")
openai.api_key = os.getenv("OPENAI_API_KEY")
openai.api_version = os.getenv('OPENAI_API_VERSION')

def get_token():
    tenantId = os.getenv("DB_TENANT_ID")
    clientId = os.getenv("DB_CLIENT_ID")
    clientSecret = os.getenv("DB_CLIENT_SECRET")

    authorityHostUrl = "https://login.microsoftonline.com"
    authorityUrl = authorityHostUrl + "/" + tenantId
    context = adal.AuthenticationContext(authorityUrl, api_version=None)
    token = context.acquire_token_with_client_credentials("https://database.windows.net/", clientId, clientSecret)

    tokenb = bytes(token["accessToken"], "UTF-8")
    exptoken = b''
    for i in tokenb:
        exptoken += bytes({i})
        exptoken += bytes(1)
    return struct.pack("=i", len(exptoken)) + exptoken

def get_conn_url():
    server = "sql-ae-dsdev-dna-hack01-pnpdebzuohv7y.database.windows.net"
    database = 'dna-hack-db01'
    return f"mssql+pyodbc://@{server}/{database}?driver=ODBC+Driver+17+for+SQL+Server"

def get_database():
    # connection_string = "mssql+pyodbc://@my-server.database.windows.net/myDb?driver=ODBC+Driver+17+for+SQL+Server"
    conn_url = get_conn_url()
    print(conn_url)
    engine = create_engine(conn_url)

    @event.listens_for(engine, "do_connect")
    def provide_token(dialect, conn_rec, cargs, cparams):
        # remove the "Trusted_Connection" parameter that SQLAlchemy adds
        cargs[0] = cargs[0].replace(";Trusted_Connection=Yes", "")

        # create token credential
        token_struct = get_token()

        # apply it to keyword arguments
        SQL_COPT_SS_ACCESS_TOKEN = 1256
        cparams["attrs_before"] = {SQL_COPT_SS_ACCESS_TOKEN: token_struct}

    return SQLDatabase(engine = engine, schema="[dntmatrix]")

def get_llm():
    return AzureChatOpenAI(temperature=1.0, 
        model_name='gpt-4',
        deployment_name='gpt-4',
        model_kwargs={
            "engine": "gpt-4",
            "api_key": openai.api_key,
            "api_base": openai.api_base,
            "api_type": openai.api_type,
            "api_version": openai.api_version
        }
    )

def get_sql_chain():
    db = get_database()
    llm = get_llm()
    return SQLDatabaseChain.from_llm(llm, db, verbose=True)

def get_sql_agent():
    db = get_database()
    llm = get_llm()
    format_instruction = """

    """ + FORMAT_INSTRUCTIONS

    toolkit = SQLDatabaseToolkit(db=db, llm=llm)
    return create_sql_agent(
        llm=llm,
        toolkit=toolkit,
        verbose=True,
        agent_type=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
        format_instructions=format_instruction
    )

sql_agent = get_sql_agent()

def predict(message, history):
    gpt_response = sql_agent.run(message)
    return gpt_response

with gr.Blocks() as demo:
    gr.ChatInterface(predict)

if __name__ == "__main__":
    demo.launch(server_name="0.0.0.0")