Upload 5 files

app.py

@@ -0,0 +1,125 @@
+import pandas as pd
+import sqlite3
+import streamlit as st
+import tempfile
+from model_api import query_hf_model  # Hugging Face API wrapper
+
+# -----------------------------
+# Connect to SQLite
+# -----------------------------
+@st.cache_resource
+def connect_sqlite(db_path="data.db"):
+    try:
+        conn = sqlite3.connect(db_path, check_same_thread=False)
+        return conn
+    except Exception as e:
+        st.error(f"Error connecting to SQLite: {e}")
+        return None
+
+# -----------------------------
+# Load CSV to SQLite
+# -----------------------------
+def load_csv_to_sqlite(file, conn):
+    try:
+        df = pd.read_csv(file)
+        df.to_sql("csv_data", conn, if_exists="replace", index=False)
+        st.success("CSV data loaded into SQLite successfully.")
+    except Exception as e:
+        st.error(f"Error loading CSV: {e}")
+
+# -----------------------------
+# Generate SQL query using HF API
+# -----------------------------
+def generate_query(user_input, conn):
+    try:
+        # Get column names from SQLite
+        cursor = conn.cursor()
+        cursor.execute("PRAGMA table_info(csv_data)")
+        field_names = [row[1] for row in cursor.fetchall()]
+
+        # Build prompt for HF model
+        prompt = f"""
+You are a MySQL expert. Only respond with a MySQL SELECT query in this exact format:
+SELECT column1, column2 FROM csv_data WHERE condition;
+
+Rules:
+- Use only these fields from the 'csv_data' table: {field_names}
+- All field names are case-sensitive.
+- String values must be in single quotes.
+- For GROUP BY queries, do not include non-aggregated columns in SELECT unless they are also in GROUP BY.
+- If extracting year from a date column (e.g., LaunchDate), use the YEAR() function in MySQL.
+- Assume all dates are stored as TEXT in the format 'YYYY-MM-DD'.
+
+User request:
+\"{user_input}\"
+"""
+
+        # Read HF token from Streamlit secrets
+        hf_token = st.secrets["HF_TOKEN"]
+        if not hf_token:
+            st.error("HF_TOKEN not found. Please add it in Streamlit secrets.")
+            return None
+
+        # Call Hugging Face API
+        query = query_hf_model(prompt, hf_token)
+
+        # Safety check: only allow SELECT queries
+        if not query.lower().strip().startswith("select"):
+            st.error("Generated SQL is not a SELECT query. Aborting.")
+            return None
+
+        return query
+    except Exception as e:
+        st.error(f"Error generating query: {e}")
+        return None
+
+# -----------------------------
+# Execute SQL query on SQLite
+# -----------------------------
+def execute_query(query, conn):
+    try:
+        df = pd.read_sql_query(query, conn)
+        return df
+    except Exception as e:
+        st.error(f"Query execution error: {e}")
+        return pd.DataFrame()
+
+# -----------------------------
+# Streamlit app
+# -----------------------------
+def main():
+    st.title("Natural Language to SQL Query and Output Generator (SQLite + HF API)")
+
+    # Connect to SQLite
+    conn = connect_sqlite()
+    if not conn:
+        st.stop()
+
+    # Upload CSV
+    uploaded_file = st.file_uploader("Upload your CSV file", type=["csv"])
+    if uploaded_file:
+        with tempfile.NamedTemporaryFile(delete=False) as tmp:
+            tmp.write(uploaded_file.read())
+            tmp_path = tmp.name
+        load_csv_to_sqlite(tmp_path, conn)
+
+    st.markdown("---")
+    user_input = st.text_input("Ask your query (in plain English):")
+    if user_input:
+        query = generate_query(user_input, conn)
+        if query:
+            st.code(query, language="sql")
+            data = execute_query(query, conn)
+            if not data.empty:
+                st.dataframe(data)
+
+                # Download result as CSV
+                csv = data.to_csv(index=False).encode("utf-8")
+                st.download_button("Download Result as CSV", csv, "result.csv", "text/csv")
+            else:
+                st.warning("No matching records found.")
+        else:
+            st.error("Could not generate a valid SQL query.")
+
+if __name__ == "__main__":
+    main()

model_api.py

@@ -0,0 +1,46 @@
+# model_api.py
+import requests
+import re
+
+HF_MODEL = "mistralai/Mistral-7B-Instruct-v0.2"  # You can change to any hosted HF model
+
+def query_hf_model(prompt: str, hf_token: str, max_tokens: int = 256):
+    """
+    Calls the Hugging Face Inference API to generate SQL from a prompt.
+    Returns the SQL query as a string.
+    """
+    api_url = f"https://api-inference.huggingface.co/models/{HF_MODEL}"
+    headers = {"Authorization": f"Bearer {hf_token}"}
+    payload = {
+        "inputs": prompt,
+        "parameters": {"max_new_tokens": max_tokens, "temperature": 0.0},
+        "options": {"wait_for_model": True}
+    }
+
+    resp = requests.post(api_url, headers=headers, json=payload, timeout=60)
+    resp.raise_for_status()
+    data = resp.json()
+
+    # Extract generated text safely
+    if isinstance(data, list) and len(data) > 0:
+        item = data[0]
+        if isinstance(item, dict):
+            text = item.get("generated_text") or item.get("text") or str(item)
+        else:
+            text = str(item)
+    elif isinstance(data, dict):
+        if "error" in data:
+            raise RuntimeError(f"Model error: {data['error']}")
+        text = data.get("generated_text") or data.get("text") or str(data)
+    else:
+        text = str(data)
+
+    # Remove code fences if present
+    text = re.sub(r"```.*?```", "", text, flags=re.S).strip()
+
+    # Return only SELECT queries
+    match = re.search(r"(?i)^\s*select\b.*", text, flags=re.S)
+    if match:
+        return match.group(0)
+    else:
+        return text.strip()

questions.txt

@@ -0,0 +1,32 @@
+Basic Level
+Select all product details.
+
+List all products in the 'Electronics' category.
+
+Get the names and prices of all products with price > 100.
+
+Count how many products are available in the dataset.
+
+List all distinct product categories.
+
+🧩 Medium Level
+Find the average rating for each category.
+
+List the top 3 most expensive products.
+
+Find the product with the highest number of reviews.
+
+Count the number of products launched in each year.
+
+List products with discounts greater than 10%.
+
+🧠 Advanced Level
+Find the category with the highest average price.
+
+List brands with more than 1 product and their average rating.
+
+Find products where stock is below the average stock for their category.
+
+Convert LaunchDate to date format and list products launched after Jan 1, 2022.
+
+Rank products within each category based on rating and review count.

requirements.txt

@@ -1,3 +1,7 @@
-altair
-pandas
-streamlit
+streamlit
+pandas
+mysql-connector-python
+langchain
+langchain-community
+langchain-ollama
+ollama

sample_data.csv

@@ -0,0 +1,11 @@
+ProductID,ProductName,Category,Price,Rating,ReviewCount,Stock,Discount,Brand,LaunchDate
+101,Wireless Mouse,Electronics,25.99,4.5,200,150,10%,Logitech,15-01-2022
+102,Gaming Keyboard,Electronics,75.49,4.7,350,85,5%,Corsair,20-11-2021
+103,Noise Cancelling Headphones,Electronics,199.99,4.8,125,60,15%,Sony,05-09-2021
+104,Running Shoes,Sports,49.99,4.3,500,200,20%,Nike,10-02-2022
+105,Smartwatch,Electronics,159.99,4.6,220,45,10%,Apple,30-03-2022
+106,Blender,Home & Kitchen,39.99,4.2,180,120,12%,Ninja,25-12-2021
+107,Coffee Maker,Home & Kitchen,89.99,4.1,240,90,8%,Keurig,15-08-2021
+108,Yoga Mat,Sports,29.99,4.4,320,300,5%,Manduka,05-01-2022
+109,Smart TV,Electronics,399.99,4.7,150,25,15%,Samsung,10-06-2022
+110,Vacuum Cleaner,Home & Kitchen,129.99,4.3,290,70,10%,Dyson,20-10-2021