Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -3,77 +3,66 @@
|
|
3 |
import gradio as gr
|
4 |
import pandas as pd
|
5 |
import torch
|
|
|
6 |
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
|
7 |
|
8 |
-
# 1) Load
|
9 |
df = pd.read_csv('synthetic_profit.csv')
|
|
|
|
|
10 |
|
11 |
-
# 2)
|
12 |
-
|
13 |
-
df[col] = pd.to_numeric(df[col], errors='coerce')
|
14 |
|
15 |
-
# 3)
|
16 |
-
# ← replaced .iteritems() with .items() here
|
17 |
-
schema_lines = [f"- {col}: {dtype.name}" for col, dtype in df.dtypes.items()]
|
18 |
-
schema_text = "Table schema:\n" + "\n".join(schema_lines)
|
19 |
-
|
20 |
-
# 4) Few-shot examples teaching SUM and AVERAGE patterns
|
21 |
-
example_block = """
|
22 |
-
Example 1
|
23 |
-
Q: Total profit by region?
|
24 |
-
A: Group “Profit” by “Region” and sum → EMEA: 30172183.37; APAC: 32301788.32; Latin America: 27585378.50; North America: 25473893.34
|
25 |
-
|
26 |
-
Example 2
|
27 |
-
Q: Average profit margin for Product B in Americas?
|
28 |
-
A: Filter Product=B & Region=Americas, take mean of “ProfitMargin” → 0.18
|
29 |
-
""".strip()
|
30 |
-
|
31 |
-
# 5) Model & pipeline setup
|
32 |
MODEL_ID = "microsoft/tapex-base-finetuned-wikisql"
|
33 |
device = 0 if torch.cuda.is_available() else -1
|
34 |
|
35 |
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
|
36 |
model = AutoModelForSeq2SeqLM.from_pretrained(MODEL_ID)
|
37 |
|
38 |
-
|
39 |
-
"
|
40 |
model=model,
|
41 |
tokenizer=tokenizer,
|
42 |
framework="pt",
|
43 |
device=device,
|
|
|
|
|
44 |
)
|
45 |
|
46 |
-
#
|
47 |
def answer_profitability(question: str) -> str:
|
48 |
-
#
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
Q: {question}
|
57 |
-
A:"""
|
58 |
|
|
|
59 |
try:
|
60 |
-
|
61 |
-
|
|
|
|
|
|
|
62 |
except Exception as e:
|
63 |
-
|
|
|
64 |
|
65 |
-
#
|
66 |
iface = gr.Interface(
|
67 |
fn=answer_profitability,
|
68 |
-
inputs=gr.Textbox(lines=2, placeholder="Ask
|
69 |
-
outputs="
|
70 |
-
title="SAP Profitability Q&A (
|
71 |
description=(
|
72 |
-
"
|
73 |
-
"
|
74 |
)
|
75 |
)
|
76 |
|
77 |
-
# 8) Launch the app
|
78 |
if __name__ == "__main__":
|
79 |
iface.launch(server_name="0.0.0.0", server_port=7860)
|
|
|
3 |
import gradio as gr
|
4 |
import pandas as pd
|
5 |
import torch
|
6 |
+
import duckdb
|
7 |
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
|
8 |
|
9 |
+
# 1) Load data and register it in DuckDB
|
10 |
df = pd.read_csv('synthetic_profit.csv')
|
11 |
+
conn = duckdb.connect(database=':memory:')
|
12 |
+
conn.register('sap', df)
|
13 |
|
14 |
+
# 2) Build a one-line schema description
|
15 |
+
schema = ", ".join(df.columns) # e.g. "Region, Product, FiscalYear, ..."
|
|
|
16 |
|
17 |
+
# 3) Load TAPEX (WikiSQL) for SQL generation
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
18 |
MODEL_ID = "microsoft/tapex-base-finetuned-wikisql"
|
19 |
device = 0 if torch.cuda.is_available() else -1
|
20 |
|
21 |
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
|
22 |
model = AutoModelForSeq2SeqLM.from_pretrained(MODEL_ID)
|
23 |
|
24 |
+
sql_generator = pipeline(
|
25 |
+
"text2text-generation",
|
26 |
model=model,
|
27 |
tokenizer=tokenizer,
|
28 |
framework="pt",
|
29 |
device=device,
|
30 |
+
# limit length so it doesn’t try to output the entire table!
|
31 |
+
max_length=128,
|
32 |
)
|
33 |
|
34 |
+
# 4) Your new QA function
|
35 |
def answer_profitability(question: str) -> str:
|
36 |
+
# 4a) Prompt the model to generate SQL
|
37 |
+
prompt = (
|
38 |
+
f"Translate to SQL for table `sap` with columns ({schema}):\n"
|
39 |
+
f"Question: {question}\n"
|
40 |
+
"SQL:"
|
41 |
+
)
|
42 |
+
sql = sql_generator(prompt)[0]['generated_text'].strip()
|
|
|
|
|
|
|
43 |
|
44 |
+
# 4b) Execute the generated SQL and return results
|
45 |
try:
|
46 |
+
result_df = conn.execute(sql).df()
|
47 |
+
# pretty-print as text
|
48 |
+
if result_df.empty:
|
49 |
+
return f"No rows returned. Generated SQL was:\n{sql}"
|
50 |
+
return result_df.to_string(index=False)
|
51 |
except Exception as e:
|
52 |
+
# if something goes wrong, show you the SQL so you can debug
|
53 |
+
return f"Error executing SQL: {e}\n\nGenerated SQL:\n{sql}"
|
54 |
|
55 |
+
# 5) Gradio interface
|
56 |
iface = gr.Interface(
|
57 |
fn=answer_profitability,
|
58 |
+
inputs=gr.Textbox(lines=2, placeholder="Ask about your SAP data…"),
|
59 |
+
outputs="textbox",
|
60 |
+
title="SAP Profitability Q&A (SQL-Generation)",
|
61 |
description=(
|
62 |
+
"Uses TAPEX to translate your natural-language question "
|
63 |
+
"into a SQL query over the `sap` table, then runs it via DuckDB."
|
64 |
)
|
65 |
)
|
66 |
|
|
|
67 |
if __name__ == "__main__":
|
68 |
iface.launch(server_name="0.0.0.0", server_port=7860)
|