Spaces:
Sleeping
Sleeping
Nam Fam commited on
Commit ·
9016439
1
Parent(s): 472e1d4
update files
Browse files- agents/sql_agent/graph.py +2 -2
- agents/sql_agent/nodes.py +178 -2
- app.py +9 -3
agents/sql_agent/graph.py
CHANGED
|
@@ -37,8 +37,8 @@ def build_graph(visualize: bool = True) -> StateGraph:
|
|
| 37 |
"detect_off_topic",
|
| 38 |
lambda state: state['error'],
|
| 39 |
path_map={
|
| 40 |
-
|
| 41 |
-
True: "get_db_info",
|
| 42 |
False: "get_db_info"
|
| 43 |
}
|
| 44 |
)
|
|
|
|
| 37 |
"detect_off_topic",
|
| 38 |
lambda state: state['error'],
|
| 39 |
path_map={
|
| 40 |
+
True: "generate_answer",
|
| 41 |
+
# True: "get_db_info",
|
| 42 |
False: "get_db_info"
|
| 43 |
}
|
| 44 |
)
|
agents/sql_agent/nodes.py
CHANGED
|
@@ -332,6 +332,180 @@ def render_visualization(state: SQLAgentState) -> SQLAgentState:
|
|
| 332 |
return state
|
| 333 |
|
| 334 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 335 |
def finalize_output(state: SQLAgentState) -> SQLAgentState:
|
| 336 |
"""
|
| 337 |
Node hợp nhất kết quả cuối cùng (answer, visualization_output, error, ...).
|
|
@@ -372,7 +546,8 @@ def detect_off_topic(state: SQLAgentState) -> SQLAgentState:
|
|
| 372 |
)
|
| 373 |
metadata = {
|
| 374 |
"topic": "Database Queries",
|
| 375 |
-
"additional_context": "
|
|
|
|
| 376 |
}
|
| 377 |
|
| 378 |
validation_result = validator.validate(question, metadata=metadata)
|
|
@@ -461,7 +636,8 @@ def generate_answer(state: SQLAgentState) -> SQLAgentState:
|
|
| 461 |
state['error'] = state['error'] or "No results found."
|
| 462 |
if state["off_topic"] == "OFF_TOPIC":
|
| 463 |
state['error'] = "The question is off-topic."
|
| 464 |
-
state["answer"] = "Sorry, I can't assist you with that request."
|
|
|
|
| 465 |
state['step'] = 'generate_answer'
|
| 466 |
return state
|
| 467 |
|
|
|
|
| 332 |
return state
|
| 333 |
|
| 334 |
|
| 335 |
+
def render_visualization(state: SQLAgentState) -> SQLAgentState:
|
| 336 |
+
"""
|
| 337 |
+
Render the visualization from formatted data.
|
| 338 |
+
Output: path to saved image file.
|
| 339 |
+
"""
|
| 340 |
+
import matplotlib.pyplot as plt
|
| 341 |
+
import os
|
| 342 |
+
import uuid
|
| 343 |
+
from typing import Dict, Any, Optional
|
| 344 |
+
|
| 345 |
+
def save_fig(fig: plt.Figure) -> str:
|
| 346 |
+
"""Save figure to file and return the file path."""
|
| 347 |
+
try:
|
| 348 |
+
output_dir = "output/plots"
|
| 349 |
+
os.makedirs(output_dir, exist_ok=True)
|
| 350 |
+
file_path = os.path.join(output_dir, f"visualization_{uuid.uuid4().hex[:8]}.png")
|
| 351 |
+
fig.savefig(file_path, format="png", bbox_inches="tight", dpi=100)
|
| 352 |
+
plt.close(fig)
|
| 353 |
+
return file_path
|
| 354 |
+
except Exception as e:
|
| 355 |
+
print(f"Error saving figure: {e}")
|
| 356 |
+
return ""
|
| 357 |
+
|
| 358 |
+
def validate_data(data: Dict[str, Any], required_keys: list) -> bool:
|
| 359 |
+
"""Validate that data contains all required keys and has valid values."""
|
| 360 |
+
if not all(key in data for key in required_keys):
|
| 361 |
+
return False
|
| 362 |
+
# Check if there's actual data to plot
|
| 363 |
+
if "values" in data and not data["values"]:
|
| 364 |
+
return False
|
| 365 |
+
if "yValues" in data and not data["yValues"]:
|
| 366 |
+
return False
|
| 367 |
+
return True
|
| 368 |
+
|
| 369 |
+
def render_line(data: Dict[str, Any]) -> Optional[str]:
|
| 370 |
+
"""Render line chart."""
|
| 371 |
+
required_keys = ["xValues", "yValues"]
|
| 372 |
+
if not validate_data(data, required_keys):
|
| 373 |
+
return None
|
| 374 |
+
|
| 375 |
+
try:
|
| 376 |
+
fig, ax = plt.subplots(figsize=(10, 6))
|
| 377 |
+
x = data["xValues"]
|
| 378 |
+
for series in data["yValues"]:
|
| 379 |
+
if len(x) == len(series["data"]):
|
| 380 |
+
ax.plot(x, series["data"], label=series.get("label", ""), marker='o')
|
| 381 |
+
|
| 382 |
+
ax.set_xlabel(data.get("xAxisLabel", "X"))
|
| 383 |
+
ax.set_ylabel(data.get("yAxisLabel", "Y"))
|
| 384 |
+
ax.set_title(data.get("title", ""))
|
| 385 |
+
if any(series.get("label") for series in data["yValues"]):
|
| 386 |
+
ax.legend()
|
| 387 |
+
plt.tight_layout()
|
| 388 |
+
return save_fig(fig)
|
| 389 |
+
except Exception as e:
|
| 390 |
+
print(f"Error rendering line chart: {e}")
|
| 391 |
+
return None
|
| 392 |
+
|
| 393 |
+
def render_bar(data: Dict[str, Any], horizontal: bool = False) -> Optional[str]:
|
| 394 |
+
"""Render bar chart (vertical or horizontal)."""
|
| 395 |
+
required_keys = ["labels", "values"]
|
| 396 |
+
if not validate_data(data, required_keys) or not data["values"]:
|
| 397 |
+
return None
|
| 398 |
+
|
| 399 |
+
try:
|
| 400 |
+
fig, ax = plt.subplots(figsize=(10, 6))
|
| 401 |
+
labels = data["labels"]
|
| 402 |
+
n_series = len(data["values"])
|
| 403 |
+
width = 0.8 / max(1, n_series) # Prevent division by zero
|
| 404 |
+
x_indexes = list(range(len(labels)))
|
| 405 |
+
|
| 406 |
+
for i, series in enumerate(data["values"]):
|
| 407 |
+
if not series["data"]: # Skip empty series
|
| 408 |
+
continue
|
| 409 |
+
|
| 410 |
+
offset = (i - n_series / 2) * width + width / 2
|
| 411 |
+
if horizontal:
|
| 412 |
+
ax.barh(
|
| 413 |
+
[x + offset for x in x_indexes],
|
| 414 |
+
series["data"],
|
| 415 |
+
height=width,
|
| 416 |
+
label=series.get("label", f"Series {i+1}")
|
| 417 |
+
)
|
| 418 |
+
ax.set_yticks(x_indexes)
|
| 419 |
+
ax.set_yticklabels(labels)
|
| 420 |
+
ax.set_xlabel(data.get("xAxisLabel", "Value"))
|
| 421 |
+
ax.set_ylabel(data.get("yAxisLabel", "Category"))
|
| 422 |
+
else:
|
| 423 |
+
ax.bar(
|
| 424 |
+
[x + offset for x in x_indexes],
|
| 425 |
+
series["data"],
|
| 426 |
+
width=width,
|
| 427 |
+
label=series.get("label", f"Series {i+1}")
|
| 428 |
+
)
|
| 429 |
+
ax.set_xticks(x_indexes)
|
| 430 |
+
ax.set_xticklabels(labels, rotation=45, ha='right')
|
| 431 |
+
ax.set_xlabel(data.get("xAxisLabel", "Category"))
|
| 432 |
+
ax.set_ylabel(data.get("yAxisLabel", "Value"))
|
| 433 |
+
|
| 434 |
+
if any(series.get("label") for series in data["values"]):
|
| 435 |
+
ax.legend()
|
| 436 |
+
ax.set_title(data.get("title", ""))
|
| 437 |
+
plt.tight_layout()
|
| 438 |
+
return save_fig(fig)
|
| 439 |
+
except Exception as e:
|
| 440 |
+
print(f"Error rendering {'horizontal ' if horizontal else ''}bar chart: {e}")
|
| 441 |
+
return None
|
| 442 |
+
|
| 443 |
+
def render_scatter(data: Dict[str, Any]) -> Optional[str]:
|
| 444 |
+
"""Render scatter plot."""
|
| 445 |
+
required_keys = ["series"]
|
| 446 |
+
if not validate_data(data, required_keys):
|
| 447 |
+
return None
|
| 448 |
+
|
| 449 |
+
try:
|
| 450 |
+
fig, ax = plt.subplots(figsize=(10, 6))
|
| 451 |
+
for series in data["series"]:
|
| 452 |
+
if not series.get("data"):
|
| 453 |
+
continue
|
| 454 |
+
xs = [point.get("x", 0) for point in series["data"]]
|
| 455 |
+
ys = [point.get("y", 0) for point in series["data"]]
|
| 456 |
+
if len(xs) == len(ys):
|
| 457 |
+
ax.scatter(
|
| 458 |
+
xs,
|
| 459 |
+
ys,
|
| 460 |
+
label=series.get("label"),
|
| 461 |
+
alpha=0.6,
|
| 462 |
+
edgecolors='w'
|
| 463 |
+
)
|
| 464 |
+
|
| 465 |
+
ax.set_xlabel(data.get("xAxisLabel", "X"))
|
| 466 |
+
ax.set_ylabel(data.get("yAxisLabel", "Y"))
|
| 467 |
+
ax.set_title(data.get("title", ""))
|
| 468 |
+
if any(series.get("label") for series in data["series"]):
|
| 469 |
+
ax.legend()
|
| 470 |
+
plt.tight_layout()
|
| 471 |
+
return save_fig(fig)
|
| 472 |
+
except Exception as e:
|
| 473 |
+
print(f"Error rendering scatter plot: {e}")
|
| 474 |
+
return None
|
| 475 |
+
|
| 476 |
+
# Main function logic
|
| 477 |
+
data = state.get("formatted_data_for_visualization")
|
| 478 |
+
visualization = state.get("visualization", "none")
|
| 479 |
+
state["visualization_output"] = None
|
| 480 |
+
|
| 481 |
+
if not data or visualization == "none":
|
| 482 |
+
return state
|
| 483 |
+
|
| 484 |
+
try:
|
| 485 |
+
renderers = {
|
| 486 |
+
"line": lambda: render_line(data),
|
| 487 |
+
"bar": lambda: render_bar(data, horizontal=False),
|
| 488 |
+
"horizontal_bar": lambda: render_bar(data, horizontal=True),
|
| 489 |
+
"scatter": lambda: render_scatter(data)
|
| 490 |
+
}
|
| 491 |
+
|
| 492 |
+
if visualization in renderers:
|
| 493 |
+
image_path = renderers[visualization]()
|
| 494 |
+
if image_path and os.path.exists(image_path):
|
| 495 |
+
state["visualization_output"] = image_path
|
| 496 |
+
else:
|
| 497 |
+
state["error"] = "Failed to generate visualization: No valid data to display"
|
| 498 |
+
else:
|
| 499 |
+
state["error"] = f"Unsupported visualization type: {visualization}"
|
| 500 |
+
|
| 501 |
+
except Exception as e:
|
| 502 |
+
state["error"] = f"Error in visualization: {str(e)}"
|
| 503 |
+
print(f"Visualization error: {e}")
|
| 504 |
+
|
| 505 |
+
state["step"] = "render_visualization"
|
| 506 |
+
return state
|
| 507 |
+
|
| 508 |
+
|
| 509 |
def finalize_output(state: SQLAgentState) -> SQLAgentState:
|
| 510 |
"""
|
| 511 |
Node hợp nhất kết quả cuối cùng (answer, visualization_output, error, ...).
|
|
|
|
| 546 |
)
|
| 547 |
metadata = {
|
| 548 |
"topic": "Database Queries",
|
| 549 |
+
"additional_context": "Only accept queries related to the data on Database/CSV"
|
| 550 |
+
# "additional_context": "The database is about ecommerce products with tables: products, laptops, phones, tablets, promotions, category"
|
| 551 |
}
|
| 552 |
|
| 553 |
validation_result = validator.validate(question, metadata=metadata)
|
|
|
|
| 636 |
state['error'] = state['error'] or "No results found."
|
| 637 |
if state["off_topic"] == "OFF_TOPIC":
|
| 638 |
state['error'] = "The question is off-topic."
|
| 639 |
+
# state["answer"] = "Sorry, I can't assist you with that request."
|
| 640 |
+
state["answer"] = "Sorry, I can only help you with questions about the data! What information would you like to explore from the data?"
|
| 641 |
state['step'] = 'generate_answer'
|
| 642 |
return state
|
| 643 |
|
app.py
CHANGED
|
@@ -464,11 +464,17 @@ else:
|
|
| 464 |
st.write(step_details.get('answer', ''))
|
| 465 |
chat_history.append({"role": "assistant", "content": step_details.get('answer', ''), "timestamp": datetime.now()})
|
| 466 |
elif step_name == 'render_visualization':
|
| 467 |
-
|
| 468 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 469 |
except Exception as e:
|
| 470 |
err = f"SQL Agent Error: {e}"
|
| 471 |
-
|
| 472 |
chat_history.append({"role": "assistant", "content": err, "timestamp": datetime.now()})
|
| 473 |
else:
|
| 474 |
# Use DataFrame agent for selected CSV
|
|
|
|
| 464 |
st.write(step_details.get('answer', ''))
|
| 465 |
chat_history.append({"role": "assistant", "content": step_details.get('answer', ''), "timestamp": datetime.now()})
|
| 466 |
elif step_name == 'render_visualization':
|
| 467 |
+
try:
|
| 468 |
+
visualization_output = step_details.get('visualization_output')
|
| 469 |
+
if visualization_output and os.path.exists(visualization_output):
|
| 470 |
+
st.image(visualization_output)
|
| 471 |
+
else:
|
| 472 |
+
print("No visualization was generated for this query.")
|
| 473 |
+
except Exception as e:
|
| 474 |
+
print(f"Could not display visualization: {str(e)}")
|
| 475 |
except Exception as e:
|
| 476 |
err = f"SQL Agent Error: {e}"
|
| 477 |
+
print(err)
|
| 478 |
chat_history.append({"role": "assistant", "content": err, "timestamp": datetime.now()})
|
| 479 |
else:
|
| 480 |
# Use DataFrame agent for selected CSV
|