app.py

import streamlit as st
import numpy as np
import pandas as pd
import altair as alt
import chat as idf_chat
from langchain.sql_database import SQLDatabase
from langchain.agents.agent_toolkits import SQLDatabaseToolkit
from langchain.agents import create_sql_agent
from langchain import OpenAI
from langchain import PromptTemplate, OpenAI, LLMChain
from langchain.chains import SimpleSequentialChain
from langchain_experimental.sql import SQLDatabaseChain
from langchain_experimental.agents.agent_toolkits import create_python_agent
from langchain_experimental.tools import PythonREPLTool
from langchain_experimental.utilities import PythonREPL


JSON_DATA_LABEL = 'json_data'
llm=OpenAI(temperature=0)
db = SQLDatabase.from_uri("sqlite:///FXTrades.db")


if JSON_DATA_LABEL not in st.session_state:
    st.session_state[JSON_DATA_LABEL] = {}


def get_db_chain():

    template = """Your name is IDF. If you recieve a "Hey IDF" salute reply by saying, "Hey BBHer!". Given an input question, first create a syntactically correct {dialect} query to run, then look at the results of the query and return the answer.
    Use the following format:
    Question: "Question here"
    SQLQuery: "SQL Query to run"
    SQLResult: "Result of the SQLQuery"
    Answer: "Final answer here"

    Only use the following tables:

    {table_info}

    If someone asks any question involving client name, you need to join with Client table
    volume: you need to count records
    Trades: you need to get volume of trades
    Currency Bought: you need to use ccyBought
    Currency Sold: you need to use ccySold


    Question: {input}"""
    PROMPT = PromptTemplate(
        input_variables=["input", "table_info", "dialect"],
        template=template
        )

    return SQLDatabaseChain.from_llm(llm, db, prompt=PROMPT, verbose=True)


def get_json_chain():
    prompt_template = "Reformat this {result} in JSON format"
    return LLMChain(
        llm=llm,
        prompt=PromptTemplate.from_template(prompt_template)
    )

def plot_chart():
    json = st.session_state[JSON_DATA_LABEL]
    # print(json)
#    if not json:
#        return "no data to plot"

    agent_executor = create_python_agent(
    llm=llm,
    tool=PythonREPLTool(),
    verbose=True
    )
    question = "Plot these results: " + json
#   Add Open AI call to format outcome in a table
    return agent_executor.run(question)

   
db_chain = get_db_chain()
json_chain = get_json_chain()

def from_gpt(query: str, plot: bool):
    try:
        chains = [db_chain, json_chain] if plot else [db_chain]
        main_chain = SimpleSequentialChain(chains=chains, verbose=True)
        ans = main_chain.run(query)
       # Execute Python if plot
        st.session_state[JSON_DATA_LABEL] = ans if plot else {}
        return ans
    except Exception as e:
        return "Please ask a proper business question related to selected datasets" 

def circle_chart():
    df = pd.DataFrame(np.random.randn(200, 3), columns=['a', 'b', 'c'])
    return alt.Chart(df).mark_circle().encode(
        x='a', y='b', size='c', color='c', tooltip=['a', 'b', 'c']
    )

# Parse the prompt to pick an example and a render function
def get_response(prompt: str, *kargs):
    on_render = st.write
    response = f"Here's what you asked: '{prompt}'"

    prompt_lower = prompt.lower()
    if prompt_lower == 'line chart':
        on_render = st.line_chart
        response = np.random.randn(30, 30)
    elif prompt_lower == 'circle chart':
        on_render = st.write
        response = circle_chart()
    elif prompt_lower.startswith('json'):
        p = prompt_lower.split('json ')[1]
        on_render = st.write
        response = from_gpt(p, plot=True)
    elif prompt_lower == 'plot':
        on_render = st.write
        response = plot_chart()
    else:
        on_render = st.write
        response = from_gpt(prompt, plot=False)

    return (response, on_render)




chat = idf_chat.Chat()

sidebar_text = """
# Ask IDF



## Example prompts
Replace `{client}` with a client name. 
Hint: you can ask IDF to tell you what clients are available

```
What is the total USD Amount for client {client} for Jan 2022 
```
```
What is the average USD Amount per month for client {client} for 2022
```
```
Trades volume for {client} in May 2022
```
```
Get the total USD amount where Currency Bought is {Currency (e.g. GBP)} for {client} in Jun 2022
```
```
Get the Currency Bought/Sold with least total USD amount for {client} in Mar 2022
```
"""

with st.sidebar:
    st.markdown(sidebar_text)

# prompt = chat.get_promt("Ask IDF Anything")

chat.process(get_response, llm)