app.py

import warnings
warnings.filterwarnings("ignore")

import torch
import gradio as gr
from transformers import AutoTokenizer, AutoModelForCausalLM

torch.set_num_threads(1)

BASE_MODEL = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"

print("Loading model...")

model = AutoModelForCausalLM.from_pretrained(
    BASE_MODEL,
    torch_dtype=torch.float32
)

tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL)
model.eval()

print("Model ready")

# ─────────────────────────
# SQL FILTER
# ─────────────────────────
SQL_KEYWORDS = [
    "sql", "database", "table", "select", "insert",
    "update", "delete", "join", "group by",
    "postgres", "mysql", "sqlite", "query"
]

def is_sql_related(text):
    text = text.lower()
    return any(k in text for k in SQL_KEYWORDS)

# ─────────────────────────
# GENERATION
# ─────────────────────────
SYSTEM_PROMPT = """
You are an expert SQL generator.

Rules:
- Only respond to SQL or database related questions.
- If the question is not about SQL or databases, refuse.
- Output ONLY SQL query.
- Do not explain.
"""

def generate_sql(user_input):

    if not user_input.strip():
        return "Enter SQL question."

    # HARD GUARD
    if not is_sql_related(user_input):
        return "I only respond to SQL and database related questions. If you want, I can craft helpful database queries for you."

    prompt = f"""
{SYSTEM_PROMPT}

User request: {user_input}
SQL:
"""

    inputs = tokenizer(prompt, return_tensors="pt")

    with torch.no_grad():
        output = model.generate(
            **inputs,
            max_new_tokens=120,
            temperature=0.1,
            do_sample=False,
            pad_token_id=tokenizer.eos_token_id,
        )

    text = tokenizer.decode(output[0], skip_special_tokens=True)

    # return only SQL part
    result = text.split("SQL:")[-1].strip()

    # extra safety: remove explanations
    result = result.split("\n\n")[0]

    return result

# ─────────────────────────
# UI
# ─────────────────────────
demo = gr.Interface(
    fn=generate_sql,
    inputs=gr.Textbox(
        lines=3,
        label="SQL Question",
        placeholder="Find duplicate emails in users table"
    ),
    outputs=gr.Textbox(
        lines=8,
        label="Generated SQL"
    ),
    title="AI SQL Generator (Portfolio Project)",
    description="This model ONLY responds to SQL/database queries.",
    examples=[
        ["Find duplicate emails in users table"],
        ["Top 5 highest paid employees"],
        ["Count orders per customer last month"],
        ["Write a joke about cats"]  # will be blocked
    ],
)

demo.launch(server_name="0.0.0.0")