Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -24,6 +24,43 @@ from typing import List, Dict, Any, Optional
|
|
| 24 |
from litellm import completion
|
| 25 |
from duckduckgo_search import DDGS
|
| 26 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 27 |
# --- Constants ---
|
| 28 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
| 29 |
DEFAULT_USERNAME = "lethaq" # 默认用户名,您可以修改为您的用户名
|
|
@@ -52,6 +89,11 @@ class DuckDuckGoSearchTool:
|
|
| 52 |
except Exception as e:
|
| 53 |
print(f"DuckDuckGo search error: {e}")
|
| 54 |
return [{"title": f"Search error: {e}", "body": "", "href": ""}]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 55 |
|
| 56 |
def __call__(self, query: str, max_results: int = 8) -> Dict[str, Any]:
|
| 57 |
"""
|
|
@@ -443,18 +485,25 @@ def run_and_submit_all():
|
|
| 443 |
for item in questions_data:
|
| 444 |
task_id = item.get("task_id")
|
| 445 |
question_text = item.get("question")
|
| 446 |
-
|
| 447 |
-
|
| 448 |
-
|
| 449 |
-
|
| 450 |
-
|
| 451 |
-
|
| 452 |
-
|
| 453 |
-
|
| 454 |
-
|
| 455 |
-
|
| 456 |
-
|
| 457 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 458 |
|
| 459 |
if not answers_payload:
|
| 460 |
print("Agent did not produce any answers to submit.")
|
|
|
|
| 24 |
from litellm import completion
|
| 25 |
from duckduckgo_search import DDGS
|
| 26 |
|
| 27 |
+
# 用於獲取並解析附件
|
| 28 |
+
def fetch_and_summarize_attachment(task_id, question, files_api_url="https://agents-course-unit4-scoring.hf.space/files/"):
|
| 29 |
+
import pandas as pd, requests
|
| 30 |
+
q_lower = question.lower()
|
| 31 |
+
keywords = ["attached", "file", "excel", "csv", "spreadsheet", "audio", "mp3", "image", "picture", "python code", "code"]
|
| 32 |
+
if not any(k in q_lower for k in keywords):
|
| 33 |
+
return ""
|
| 34 |
+
try:
|
| 35 |
+
file_url = f"{files_api_url}{task_id}"
|
| 36 |
+
response = requests.get(file_url, timeout=10)
|
| 37 |
+
local_path = f"/tmp/{task_id}"
|
| 38 |
+
with open(local_path, "wb") as f:
|
| 39 |
+
f.write(response.content)
|
| 40 |
+
if local_path.endswith(".csv") or b"," in response.content[:512]:
|
| 41 |
+
df = pd.read_csv(local_path)
|
| 42 |
+
summary = df.head(10).to_string()
|
| 43 |
+
return f"[Attachment Detected: CSV Table]\nFirst 10 rows:\n{summary}\n"
|
| 44 |
+
elif local_path.endswith(".xls") or local_path.endswith(".xlsx"):
|
| 45 |
+
df = pd.read_excel(local_path)
|
| 46 |
+
summary = df.head(10).to_string()
|
| 47 |
+
return f"[Attachment Detected: Excel Table]\nFirst 10 rows:\n{summary}\n"
|
| 48 |
+
elif local_path.endswith(".txt") or b"\n" in response.content[:512]:
|
| 49 |
+
with open(local_path, "r", encoding="utf-8", errors="ignore") as f:
|
| 50 |
+
text = f.read(1000)
|
| 51 |
+
return f"[Attachment Detected: Text File]\n{text}\n"
|
| 52 |
+
elif local_path.endswith(".py"):
|
| 53 |
+
with open(local_path, "r", encoding="utf-8", errors="ignore") as f:
|
| 54 |
+
code = f.read(1000)
|
| 55 |
+
return f"[Attachment Detected: Python Code]\n{code}\n"
|
| 56 |
+
else:
|
| 57 |
+
return "[Attachment Detected, but file type not supported for summary]\n"
|
| 58 |
+
except Exception as e:
|
| 59 |
+
return f"[Attachment Error: {e}]\n"
|
| 60 |
+
|
| 61 |
+
|
| 62 |
+
|
| 63 |
+
|
| 64 |
# --- Constants ---
|
| 65 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
| 66 |
DEFAULT_USERNAME = "lethaq" # 默认用户名,您可以修改为您的用户名
|
|
|
|
| 89 |
except Exception as e:
|
| 90 |
print(f"DuckDuckGo search error: {e}")
|
| 91 |
return [{"title": f"Search error: {e}", "body": "", "href": ""}]
|
| 92 |
+
|
| 93 |
+
|
| 94 |
+
|
| 95 |
+
|
| 96 |
+
|
| 97 |
|
| 98 |
def __call__(self, query: str, max_results: int = 8) -> Dict[str, Any]:
|
| 99 |
"""
|
|
|
|
| 485 |
for item in questions_data:
|
| 486 |
task_id = item.get("task_id")
|
| 487 |
question_text = item.get("question")
|
| 488 |
+
if not task_id or question_text is None:
|
| 489 |
+
print(f"Skipping item with missing task_id or question: {item}")
|
| 490 |
+
continue
|
| 491 |
+
try:
|
| 492 |
+
# ① Attachment 摘要插入
|
| 493 |
+
attachment_context = fetch_and_summarize_attachment(task_id, question_text)
|
| 494 |
+
# ② 拼 prompt(根據你的 agent 用法):
|
| 495 |
+
# (方案A)如果 agent 支援 attachment_context 參數
|
| 496 |
+
# submitted_answer = agent(question_text, attachment_context=attachment_context)
|
| 497 |
+
# (方案B)最直接:把 attachment 拼到問題開頭
|
| 498 |
+
full_question = (attachment_context + "\n" + question_text).strip()
|
| 499 |
+
submitted_answer = agent(full_question)
|
| 500 |
+
answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
|
| 501 |
+
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
|
| 502 |
+
print(f"Answer for task {task_id}: {submitted_answer[:50]}...")
|
| 503 |
+
except Exception as e:
|
| 504 |
+
print(f"Error running agent on task {task_id}: {e}")
|
| 505 |
+
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
|
| 506 |
+
|
| 507 |
|
| 508 |
if not answers_payload:
|
| 509 |
print("Agent did not produce any answers to submit.")
|