app.py

# app.py  —  BizIntel AI Ultra  
# Supports: CSV/Excel/DB ingestion, date+metric plotting, ARIMA forecasting,
# safe Plotly writes into /tmp, Gemini 1.5 Pro strategy, KPI cards, optional EDA.

import os
import tempfile

import pandas as pd
import streamlit as st
import google.generativeai as genai
import plotly.graph_objects as go

# ──────────────────────────────────────────────────────────────
# 0) Monkey‐patch Plotly to write images into /tmp (writable)
# ──────────────────────────────────────────────────────────────
TMP = tempfile.gettempdir()
_orig_write = go.Figure.write_image
def _safe_write(self, path, *args, **kwargs):
    filename = os.path.basename(path)
    safe_path = os.path.join(TMP, filename)
    return _orig_write(self, safe_path, *args, **kwargs)
go.Figure.write_image = _safe_write

# ──────────────────────────────────────────────────────────────
# 1) Tool & DB imports
# ──────────────────────────────────────────────────────────────
from tools.csv_parser      import parse_csv_tool
from tools.plot_generator  import plot_metric_tool
from tools.forecaster      import forecast_metric_tool
from tools.visuals         import histogram_tool, scatter_matrix_tool, corr_heatmap_tool
from db_connector          import fetch_data_from_db, list_tables, SUPPORTED_ENGINES

# ──────────────────────────────────────────────────────────────
# 2) Gemini 1.5 Pro initialization
# ──────────────────────────────────────────────────────────────
genai.configure(api_key=os.getenv("GEMINI_APIKEY"))
gemini = genai.GenerativeModel(
    "gemini-1.5-pro-latest",
    generation_config={
        "temperature": 0.7,
        "top_p": 0.9,
        "response_mime_type": "text/plain",
    },
)

# ──────────────────────────────────────────────────────────────
# 3) Streamlit page setup
# ──────────────────────────────────────────────────────────────
st.set_page_config(page_title="BizIntel AI Ultra", layout="wide")
st.title("📊 BizIntel AI Ultra – Advanced Analytics + Gemini 1.5 Pro")
TEMP_DIR = tempfile.gettempdir()

# ──────────────────────────────────────────────────────────────
# 4) Data source selection: CSV/Excel or SQL Database
# ──────────────────────────────────────────────────────────────
source = st.radio("Select data source", ["Upload CSV / Excel", "Connect to SQL Database"])
csv_path = None

if source == "Upload CSV / Excel":
    upload = st.file_uploader("Upload CSV or Excel (≤ 500 MB)", type=["csv","xlsx","xls"])
    if upload:
        tmp_file = os.path.join(TEMP_DIR, upload.name)
        with open(tmp_file, "wb") as f:
            f.write(upload.read())

        if upload.name.lower().endswith(".csv"):
            csv_path = tmp_file
        else:
            try:
                df_xl = pd.read_excel(tmp_file, sheet_name=0)
                csv_path = tmp_file.rsplit(".", 1)[0] + ".csv"
                df_xl.to_csv(csv_path, index=False)
            except Exception as e:
                st.error(f"Excel parsing failed: {e}")
                st.stop()
        st.success(f"{upload.name} saved ✅")

else:
    engine = st.selectbox("DB engine", SUPPORTED_ENGINES)
    conn   = st.text_input("SQLAlchemy connection string")
    if conn:
        try:
            tables = list_tables(conn)
            tbl    = st.selectbox("Table", tables)
            if st.button("Fetch table"):
                csv_path = fetch_data_from_db(conn, tbl)
                st.success(f"Fetched **{tbl}** as CSV ✅")
        except Exception as e:
            st.error(f"Connection failed: {e}")
            st.stop()

if not csv_path:
    st.stop()

# Download the working CSV
with open(csv_path, "rb") as f:
    st.download_button("⬇️ Download working CSV", f, file_name=os.path.basename(csv_path))

# ──────────────────────────────────────────────────────────────
# 5) Show head & pick date + metric (but never the same column)
# ──────────────────────────────────────────────────────────────
df_head = pd.read_csv(csv_path, nrows=5)
st.dataframe(df_head)

# a) Date dropdown over all columns
date_col = st.selectbox("Select date/time column", df_head.columns)

# b) Metric dropdown only numeric columns, excluding the chosen date_col
numeric_cols = df_head.select_dtypes("number").columns.tolist()
metric_options = [c for c in numeric_cols if c != date_col]
if not metric_options:
    st.error(f"No numeric columns available once we exclude '{date_col}'.")
    st.stop()
metric_col = st.selectbox("Select numeric metric column", metric_options)

# ──────────────────────────────────────────────────────────────
# 6) Local analysis: summary, trend chart, forecast
# ──────────────────────────────────────────────────────────────
with st.spinner("Parsing dataset…"):
    summary_text = parse_csv_tool(csv_path)

with st.spinner("Generating trend chart…"):
    trend_fig = plot_metric_tool(csv_path, date_col, metric_col)
if isinstance(trend_fig, go.Figure):
    st.subheader("📈 Trend")
    st.plotly_chart(trend_fig, use_container_width=True)
else:
    st.warning(trend_fig)

with st.spinner("Running forecast…"):
    forecast_text = forecast_metric_tool(csv_path, date_col, metric_col)

st.subheader(f"🔮 {metric_col} Forecast")
forecast_png = os.path.join(TEMP_DIR, "forecast_plot.png")
if os.path.exists(forecast_png):
    st.image(forecast_png, use_container_width=True)
else:
    st.warning("Forecast image not found.")

# ──────────────────────────────────────────────────────────────
# 7) Gemini-driven strategy recommendations
# ──────────────────────────────────────────────────────────────
prompt = (
    f"You are **BizIntel Strategist AI**.\n\n"
    f"### Dataset Summary\n```\n{summary_text}\n```\n\n"
    f"### {metric_col} Forecast\n```\n{forecast_text}\n```\n\n"
    "Return **Markdown** with:\n"
    "1. Five key insights\n"
    "2. Three actionable strategies\n"
    "3. Risk factors or anomalies\n"
    "4. Suggested additional visuals\n"
)

st.subheader("🚀 Strategy Recommendations (Gemini 1.5 Pro)")
with st.spinner("Generating insights…"):
    strategy_md = gemini.generate_content(prompt).text
st.markdown(strategy_md)
st.download_button("⬇️ Download Strategy (.md)", strategy_md, file_name="strategy.md")

# ──────────────────────────────────────────────────────────────
# 8) KPI cards + detailed Stats
# ──────────────────────────────────────────────────────────────
full_df     = pd.read_csv(csv_path, low_memory=False)
total_rows  = len(full_df)
num_columns = full_df.shape[1]
missing_pct = full_df.isna().mean().mean() * 100

st.markdown("---")
st.subheader("📑 Dataset Overview")
c1, c2, c3 = st.columns(3)
c1.metric("Rows",      f"{total_rows:,}")
c2.metric("Columns",   str(num_columns))
c3.metric("Missing %", f"{missing_pct:.1f}%")

with st.expander("🔎 Detailed descriptive statistics"):
    stats_df = (
        full_df.describe()
        .T
        .reset_index()
        .rename(columns={"index":"Feature"})
    )
    st.dataframe(
        stats_df.style.format(precision=2).background_gradient(cmap="Blues"),
        use_container_width=True
    )

# ──────────────────────────────────────────────────────────────
# 9) Optional Exploratory Visuals
# ──────────────────────────────────────────────────────────────
st.markdown("---")
st.subheader("🔍 Optional Exploratory Visuals")

if st.checkbox("Histogram"):
    hcol = st.selectbox("Variable", numeric_cols, key="hist")
    st.plotly_chart(histogram_tool(csv_path, hcol), use_container_width=True)

if st.checkbox("Scatter Matrix"):
    sel = st.multiselect("Choose columns", numeric_cols, default=numeric_cols[:3])
    if sel:
        st.plotly_chart(scatter_matrix_tool(csv_path, sel), use_container_width=True)

if st.checkbox("Correlation Heatmap"):
    st.plotly_chart(corr_heatmap_tool(csv_path), use_container_width=True)