File size: 7,470 Bytes
c3b3775
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
import os
import tempfile

import gradio as gr
import torch
from peft import PeftConfig, PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

from schema_extractor import SQLiteSchemaExtractor


# Load model and tokenizer
def load_model():
    config = PeftConfig.from_pretrained("Rajan/training_run")
    tokenizer = AutoTokenizer.from_pretrained("NumbersStation/nsql-350M")
    base_model = AutoModelForCausalLM.from_pretrained("NumbersStation/nsql-350M")
    model = PeftModel.from_pretrained(base_model, "Rajan/training_run")
    return model, tokenizer


# Extract and correct SQL from generated text
def extract_and_correct_sql(text, correct=False):
    lines = text.splitlines()

    start_index = 0
    for i, line in enumerate(lines):
        if line.strip().upper().startswith("SELECT"):
            start_index = i
            break

    generated_sql = "\n".join(lines[start_index:])

    if correct:
        if not generated_sql.strip().endswith(";"):
            generated_sql = generated_sql.strip() + ";"

    return generated_sql


# Function to handle file upload and schema extraction
def upload_and_extract_schema(db_file):
    if db_file is None:
        return "Please upload a database file", None

    try:
        # Get the file path directly from Gradio
        temp_db_path = db_file.name

        extractor = SQLiteSchemaExtractor(temp_db_path)
        schema = extractor.get_schema()
        return schema, temp_db_path
    except Exception as e:
        return f"Error extracting schema: {str(e)}", None


# Function to handle chat interaction with streaming effect
def generate_sql(question, schema, db_path, chat_history):
    if db_path is None or not schema:
        return (
            chat_history
            + [
                {"role": "user", "content": question},
                {"role": "assistant", "content": "Please upload a database file first"},
            ],
            None,
        )

    try:
        # Load model
        model, tokenizer = load_model()

        # Format prompt
        prompt_format = """ 
{}
-- Using valid SQLite, answer the following questions for the tables provided above.
{}
SELECT"""

        # Format the prompt with schema and question
        prompt = prompt_format.format(schema, question)

        # Generate SQL
        input_ids = tokenizer(prompt, return_tensors="pt").input_ids
        generated_ids = model.generate(input_ids, max_length=500)
        generated_text = tokenizer.decode(generated_ids[0], skip_special_tokens=True)

        # Extract SQL
        sql_query = extract_and_correct_sql(generated_text, correct=True)

        # Update history using dictionary format
        new_history = chat_history + [
            {"role": "user", "content": question},
            {"role": "assistant", "content": sql_query},
        ]
        return new_history, sql_query
    except Exception as e:
        error_message = f"Error: {str(e)}"
        return (
            chat_history
            + [
                {"role": "user", "content": question},
                {"role": "assistant", "content": error_message},
            ],
            None,
        )


# Function for streaming SQL generation effect
def stream_sql(question, schema, db_path, chat_history):
    # First add the user message to chat
    yield chat_history + [{"role": "user", "content": question}], ""

    if db_path is None or not schema:
        yield chat_history + [
            {"role": "user", "content": question},
            {"role": "assistant", "content": "Please upload a database file first"},
        ], "Please upload a database file first"
        return

    try:
        # Load model
        model, tokenizer = load_model()

        # Format prompt
        prompt_format = """ 
{}
-- Using valid SQLite, answer the following questions for the tables provided above.
{}
SELECT"""

        # Format the prompt with schema and question
        prompt = prompt_format.format(schema, question)

        # Generate SQL
        input_ids = tokenizer(prompt, return_tensors="pt").input_ids
        generated_ids = model.generate(input_ids, max_length=500)
        generated_text = tokenizer.decode(generated_ids[0], skip_special_tokens=True)

        # Extract SQL
        sql_query = extract_and_correct_sql(generated_text, correct=True)

        # Fixed medium speed (0.03 seconds delay)
        import time

        delay = 0.03  # 30ms - normal typing speed

        # Stream the SQL query character by character for effect
        partial_sql = ""
        for char in sql_query:
            partial_sql += char
            # Update chat history and SQL display with each character
            yield chat_history + [
                {"role": "user", "content": question},
                {"role": "assistant", "content": partial_sql},
            ], partial_sql
            time.sleep(delay)  # Medium speed typing effect

    except Exception as e:
        error_message = f"Error: {str(e)}"
        yield chat_history + [
            {"role": "user", "content": question},
            {"role": "assistant", "content": error_message},
        ], error_message


# Main application
def create_app():
    with gr.Blocks(title="SQL Query Generator", theme=gr.themes.Soft()) as app:
        gr.Markdown("# SQL Query Generator")
        gr.Markdown(
            "Upload a SQLite database, ask questions, and get SQL queries automatically generated"
        )

        # Store database path
        db_path = gr.State(value=None)

        with gr.Row():
            with gr.Column(scale=1):
                # File upload section
                file_input = gr.File(label="Upload SQLite Database (.db file)")
                extract_btn = gr.Button("Extract Schema", variant="primary")

                # Schema display
                schema_output = gr.Textbox(
                    label="Database Schema", lines=10, interactive=False
                )

            with gr.Column(scale=2):
                # Chat interface
                chatbot = gr.Chatbot(
                    label="Query Conversation", height=400, type="messages"
                )

                with gr.Row():
                    question_input = gr.Textbox(
                        label="Ask a question about your data",
                        placeholder="e.g., Show me the top 10 most sold items",
                    )
                    submit_btn = gr.Button("Generate SQL", variant="primary")

                # SQL output display
                sql_output = gr.Code(
                    language="sql", label="Generated SQL Query", interactive=False
                )

        # Event handlers
        extract_btn.click(
            fn=upload_and_extract_schema,
            inputs=[file_input],
            outputs=[schema_output, db_path],
        )

        submit_btn.click(
            fn=stream_sql,
            inputs=[question_input, schema_output, db_path, chatbot],
            outputs=[chatbot, sql_output],
            api_name="generate",
            queue=True,
        )

        # Also trigger on enter key
        question_input.submit(
            fn=stream_sql,
            inputs=[question_input, schema_output, db_path, chatbot],
            outputs=[chatbot, sql_output],
            api_name="generate_on_submit",
            queue=True,
        )

    return app


# Launch the app
if __name__ == "__main__":
    app = create_app()
    app.launch(share=True)