src/streamlit_app.py

import os
import pandas as pd
import numpy as np
import streamlit as st
import plotly.express as px
import plotly.figure_factory as ff
from dotenv import load_dotenv
from huggingface_hub import InferenceClient, login
from io import StringIO
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline

# ======================================================
# ⚙️ APP CONFIGURATION
# ======================================================
st.set_page_config(page_title="📊 Smart Data Analyst Pro", layout="wide")
st.title("📊 Smart Data Analyst Pro")
st.caption("AI that cleans, analyzes, and visualizes your data — powered by Hugging Face Inference API and local open-source models.")

# ======================================================
# 🔐 Load Environment Variables
# ======================================================
load_dotenv()
HF_TOKEN = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACE_API_KEY")
if not HF_TOKEN:
    st.error("❌ Missing HF_TOKEN. Please set it in your .env file.")
else:
    login(token=HF_TOKEN)

# ======================================================
# 🧠 MODEL SETTINGS
# ======================================================
with st.sidebar:
    st.header("⚙️ Model Settings")

    CLEANER_MODEL = st.selectbox(
        "Select Cleaner Model:",
        [
            "Qwen/Qwen2.5-Coder-7B-Instruct",
            "meta-llama/Meta-Llama-3-8B-Instruct",
            "microsoft/Phi-3-mini-4k-instruct"
        ],
        index=0
    )

    ANALYST_MODEL = st.selectbox(
        "Select Analysis Model (Local Open-Source Recommended):",
        [            
            "meta-llama/Meta-Llama-3-8B-Instruct",
            "Qwen/Qwen2.5-Coder-7B-Instruct",
            "HuggingFaceH4/zephyr-7b-beta",
            "mistralai/Mistral-7B-Instruct-v0.3"
        ],
        index=0
    )

    temperature = st.slider("Temperature", 0.0, 1.0, 0.3)
    max_tokens = st.slider("Max Tokens", 128, 2048, 512)

# Initialize cleaner client (HF API)
cleaner_client = InferenceClient(model=CLEANER_MODEL, token=HF_TOKEN)

# Initialize local analyst if open-source
local_analyst = None
if ANALYST_MODEL in ["meta-llama/Meta-Llama-3-8B-Instruct"]:
    try:
        tokenizer = AutoTokenizer.from_pretrained(ANALYST_MODEL)
        model = AutoModelForCausalLM.from_pretrained(ANALYST_MODEL)
        local_analyst = pipeline("text-generation", model=model, tokenizer=tokenizer)
    except Exception as e:
        st.warning(f"⚠️ Failed to load local analyst: {e}")

# ======================================================
# 🧩 DATA CLEANING FUNCTIONS
# ======================================================
def fallback_clean(df: pd.DataFrame) -> pd.DataFrame:
    df = df.copy()
    df.dropna(axis=1, how="all", inplace=True)
    df.columns = [c.strip().replace(" ", "_").lower() for c in df.columns]
    for col in df.columns:
        if df[col].dtype == "O":
            df[col].fillna(df[col].mode()[0] if not df[col].mode().empty else "Unknown", inplace=True)
        else:
            df[col].fillna(df[col].median(), inplace=True)
    df.drop_duplicates(inplace=True)
    return df

def ai_clean_dataset(df: pd.DataFrame) -> pd.DataFrame:
    raw_preview = df.head(5).to_csv(index=False)
    prompt = f"""
You are a Python data cleaning expert.
Clean and standardize the dataset dynamically:
- Handle missing values logically
- Correct and normalize column names
- Detect and fix datatype inconsistencies
- Remove duplicates or invalid rows
Return ONLY valid CSV text (no Markdown).

--- RAW SAMPLE ---
{raw_preview}
"""
    try:
        response = cleaner_client.text_generation(prompt, max_new_tokens=1024, temperature=0.1, return_full_text=False)
        cleaned_str = response.strip()
    except Exception as e:
        st.warning(f"⚠️ AI cleaning failed: {e}")
        return fallback_clean(df)

    cleaned_str = cleaned_str.replace("```csv","").replace("```","").replace("###","").replace(";",",").strip()
    lines = [l for l in cleaned_str.splitlines() if "," in l]
    cleaned_str = "\n".join(lines)

    try:
        cleaned_df = pd.read_csv(StringIO(cleaned_str), on_bad_lines="skip")
        cleaned_df.dropna(axis=1, how="all", inplace=True)
        cleaned_df.columns = [c.strip().replace(" ", "_").lower() for c in cleaned_df.columns]
        return cleaned_df
    except Exception as e:
        st.warning(f"⚠️ CSV parse failed: {e}")
        return fallback_clean(df)

def summarize_dataframe(df: pd.DataFrame) -> str:
    lines = [f"Rows: {len(df)} | Columns: {len(df.columns)}", "Column summaries:"]
    for col in df.columns[:10]:
        non_null = int(df[col].notnull().sum())
        if pd.api.types.is_numeric_dtype(df[col]):
            mean = df[col].mean()
            median = df[col].median() if non_null > 0 else None
            lines.append(f"- {col}: mean={mean:.3f}, median={median}, non_null={non_null}")
        else:
            top = df[col].value_counts().head(3).to_dict()
            lines.append(f"- {col}: top_values={top}, non_null={non_null}")
    return "\n".join(lines)

# ======================================================
# 🧠 ANALYSIS FUNCTION
# ======================================================
def query_analysis_model(df: pd.DataFrame, user_query: str, dataset_name: str) -> str:
    df_summary = summarize_dataframe(df)
    sample = df.head(6).to_csv(index=False)
    prompt = f"""
You are a data analyst.
Analyze '{dataset_name}' and answer the question below.
Base your insights only on the provided data.

--- SUMMARY ---
{df_summary}

--- SAMPLE DATA ---
{sample}

--- QUESTION ---
{user_query}

Respond concisely with key insights, numbers, patterns, and recommended steps.
"""
    if local_analyst:
        try:
            response = local_analyst(prompt, max_new_tokens=max_tokens, temperature=temperature)
            return response[0]['generated_text']
        except Exception as e:
            return f"⚠️ Local analysis failed: {e}"
    else:
        st.warning("⚠️ Analyst model is not local. Using HF API may require payment.")
        return "Analysis not available for free model."

# ======================================================
# 🚀 MAIN APP
# ======================================================
uploaded = st.file_uploader("📎 Upload CSV or Excel file", type=["csv", "xlsx"])

if uploaded:
    try:
        df = pd.read_csv(uploaded) if uploaded.name.endswith(".csv") else pd.read_excel(uploaded)
    except Exception as e:
        st.error(f"❌ File load failed: {e}")
        st.stop()

    with st.spinner("🧼 AI Cleaning your dataset..."):
        cleaned_df = ai_clean_dataset(df)

    st.subheader("✅ Cleaned Dataset Preview")
    st.dataframe(cleaned_df.head(), use_container_width=True)

    with st.expander("📋 Cleaning Summary"):
        st.text(summarize_dataframe(cleaned_df))

    with st.expander("📈 Quick Visualizations", expanded=True):
        numeric_cols = cleaned_df.select_dtypes(include="number").columns.tolist()
        categorical_cols = cleaned_df.select_dtypes(exclude="number").columns.tolist()
        viz_type = st.selectbox("Visualization Type", ["Scatter Plot", "Histogram", "Box Plot", "Correlation Heatmap", "Categorical Count"])

        if viz_type == "Scatter Plot" and len(numeric_cols) >= 2:
            x = st.selectbox("X-axis", numeric_cols)
            y = st.selectbox("Y-axis", numeric_cols, index=min(1,len(numeric_cols)-1))
            color = st.selectbox("Color", ["None"] + categorical_cols)
            fig = px.scatter(cleaned_df, x=x, y=y, color=None if color=="None" else color)
            st.plotly_chart(fig, use_container_width=True)
        elif viz_type == "Histogram" and numeric_cols:
            col = st.selectbox("Column", numeric_cols)
            fig = px.histogram(cleaned_df, x=col, nbins=30)
            st.plotly_chart(fig, use_container_width=True)
        elif viz_type == "Box Plot" and numeric_cols:
            col = st.selectbox("Column", numeric_cols)
            fig = px.box(cleaned_df, y=col)
            st.plotly_chart(fig, use_container_width=True)
        elif viz_type == "Correlation Heatmap" and len(numeric_cols) > 1:
            corr = cleaned_df[numeric_cols].corr()
            fig = ff.create_annotated_heatmap(z=corr.values, x=list(corr.columns), y=list(corr.index),
                                              annotation_text=corr.round(2).values, showscale=True)
            st.plotly_chart(fig, use_container_width=True)
        elif viz_type == "Categorical Count" and categorical_cols:
            cat = st.selectbox("Category", categorical_cols)
            fig = px.bar(cleaned_df[cat].value_counts().reset_index(), x="index", y=cat)
            st.plotly_chart(fig, use_container_width=True)
        else:
            st.warning("⚠️ Not enough columns for this visualization type.")

    st.subheader("💬 Ask AI About Your Data")
    user_query = st.text_area("Enter your question:", placeholder="e.g. What factors influence sales?")
    if st.button("Analyze with AI", use_container_width=True) and user_query:
        with st.spinner("🤖 Interpreting data..."):
            result = query_analysis_model(cleaned_df, user_query, uploaded.name)
        st.markdown("### 💡 Insights")
        st.markdown(result)
else:
    st.info("📥 Upload a dataset to begin smart analysis.")