streamlit_app.py

import streamlit as st
import numpy as np
from snowflake.snowpark.context import get_active_session
from snowflake.cortex import Complete
import pandas as pd

st.set_page_config(layout="wide")


def is_local() -> bool:
    """
    Check if app is running locally, by checking if the user email is a local one.

    Returns:
        bool: True if running locally, else (if in SiS) False.
    """
    return st.experimental_user.email in {"test@localhost.com", "test@example.com"}

@st.cache_data
def generate_fake_data_for_demo(seed=36) -> pd.DataFrame:
    """
    Generates a dataframe of customers and views for
    this demo, and caches it. If you are using this demo,
    you ideally will replace this with your own data.
    """

    df = pd.DataFrame(
        {
            "date": pd.date_range(start="1/1/2020", periods=100),
            "customers": np.random.randint(100, 1000, 100),
        }
    )
    df["customers"] = df["customers"].cumsum()
    np.random.seed(seed)
    df["views"] = np.random.randint(100, 1000, 100)
    df["views"] = df["views"] * np.log(df.index) * np.random.uniform(1, 1.6, 100)
    df["views"][0] = 0

    return df


st.title(f"❄️ Streamlit in Snowflake Key Metrics (dummy data) ❄️")
st.sidebar.image(
    "https://upload.wikimedia.org/wikipedia/commons/f/ff/Snowflake_Logo.svg"
)

if is_local():
    conn = st.connection("snowflake")
    session = conn.session()

else:
    session = get_active_session()

df = generate_fake_data_for_demo()

col1, col2 = st.columns(2)
with col1:
    st.subheader("Number of customers")
    taba, tabb = st.tabs(["Chart", "Data"])
    taba.line_chart(df.set_index("date")["customers"])
    tabb.write(df)


with col2:
    st.subheader("Views per day")
    tabc, tabd = st.tabs(["Chart", "Data"])
    tabc.line_chart(df.set_index("date")["views"])
    tabd.write(df)




df["views_per_customer"] = df["views"] / df["customers"]
st.subheader("Views per customer")
st.line_chart(df.set_index("date")["views_per_customer"])

prompt = """
Please summarize the following feedback comments
    in markdown from our streamlit in snowflake users,
    just give the top 3 good things and 3 improvement areas about the product: '
"""
df_feedback = session.sql("select * from streamlit.public.feedback_table").to_pandas()
prompt += "\n\n".join(df_feedback["FEEDBACK"])
st.subheader("User Feedback")
response = Complete(model="mistral-large", prompt=prompt, session=session)
st.write(response)