Tracy / app.py
mogg7's picture
Update app.py
55260df verified
Raw History Blame
6.36 kB
from smolagents import CodeAgent, DuckDuckGoSearchTool, HfApiModel, load_tool, tool
import datetime
import requests
import pytz
import yaml
from tools.visit_webpage import VisitWebpageTool
from tools.final_answer import FinalAnswerTool
from Gradio_UI import GradioUI
from typing import List, Dict
from bs4 import BeautifulSoup
@tool
def search_terms_of_service(company_name: str) -> Dict:
"""Search for a company's Terms of Service link
Args:
company_name: Name of the organization/company to search for
Returns:
Dictionary containing company name and link to their Terms of Service
"""
try:
# API URL
# API, ki poskrbi za komunikacijo s podatkovno bazo se poganja lokalno
# Da omogočimo javni dostop do API se uporabi ngrok URL povezava, ki se ob vsakem zagonu spremeni
# Posledično je treba ngrok URL stalno posodabljati
api_base = "https://3aec-176-76-224-7.ngrok-free.app"
# API, ki poskrbi za komunikacijo s podatkovno bazo s
# API request naredimo
url = f"{api_base}/company/{company_name}"
print(f"Request URL: {url}")
response = requests.get(url)
print(f"Response status: {response.status_code}")
print(f"Response body preview: {response.text[:200]}...")
if response.status_code == 404: #HTTP error!
return {"error": f"Company '{company_name}' not found in the database"}
response.raise_for_status()
data = response.json()
# ali obstaja več linkov?
tos_link = None
for link in data['secondary_links']:
if 'terms' in link.lower():
tos_link = link
break
# If no link with "terms" found, fall back to first link
if not tos_link:
tos_link = data["secondary_links"][0]
print(f"Warning: No link containing 'terms' found, using first link: {tos_link}")
# returnamo ime in link, lahko jih več
result = {
"company_name": data["name"],
"tos_link": tos_link,
"all_links": data["secondary_links"] # več linkov
}
return result
except requests.exceptions.RequestException as e:
error_msg = f"API Error: {str(e)}"
if hasattr(e, 'response') and e.response is not None:
error_msg += f" (Status code: {e.response.status_code})"
if hasattr(e.response, 'text'):
error_msg += f" - {e.response.text[:200]}"
print(error_msg) # debug, da prizna, če se zmoti
return {"error": error_msg}
except Exception as e:
error_msg = f"Unexpected error: {str(e)}"
print(error_msg) # debug, da AI prizna, če se zmoti
return {"error": error_msg}
@tool
def extract_tos_content(company_name: str) -> Dict:
"""
Search for a company's Terms of Service, access the link, and extract raw content from GitLab repository.
Args:
company_name: Name of the organization/company to search for
Returns:
Dictionary containing company name, link, and extracted content
"""
tos_info = search_terms_of_service(company_name)
# error
if "error" in tos_info:
return tos_info
try:
# link od zgoraj
tos_link = tos_info["tos_link"]
print(f"Accessing TOS link: {tos_link}")
# Gitlab repo
if "code.europa.eu" in tos_link and "/-/blob/" in tos_link:
# spremeni '/-/blob/' z '/-/raw/'
raw_url = tos_link.replace("/-/blob/", "/-/raw/")
if "?" in raw_url:
raw_url = raw_url.split("?")[0]
print(f"Accessing raw content URL: {raw_url}")
# request za raw data iz GItlabaa
response = requests.get(raw_url)
response.raise_for_status()
return {
"company_name": tos_info["company_name"],
"tos_link": tos_link,
"raw_url": raw_url,
"content": response.text, # cel raw data fajl
"content_type": "markdown" if raw_url.endswith(".md") else "text"
}
response = requests.get(tos_link)
response.raise_for_status()
soup = BeautifulSoup(response.text, 'html.parser')
content = ""
pre_element = soup.find('pre')
if pre_element:
content = pre_element.get_text()
else:
main_content = soup.find('main') or soup.find('article') or soup.find('body')
if main_content:
content = main_content.get_text(strip=True)
return {
"company_name": tos_info["company_name"],
"tos_link": tos_link,
"content": content,
"note": "Content extracted from HTML page"
}
except requests.exceptions.RequestException as e:
error_msg = f"Error accessing TOS link: {str(e)}"
if hasattr(e, 'response') and e.response is not None:
error_msg += f" (Status code: {e.response.status_code})"
if hasattr(e.response, 'text'):
error_msg += f" - {e.response.text[:200]}"
print(error_msg)
return {"error": error_msg}
except Exception as e:
error_msg = f"Unexpected error while processing content: {str(e)}"
print(error_msg)
return {"error": error_msg}
final_answer = FinalAnswerTool()
# If the agent does not answer, the model is overloaded, please use another model or the following Hugging Face Endpoint that also contains qwen2.5 coder:
# model_id='https://pflgm2locj2t89co.us-east-1.aws.endpoints.huggingface.cloud'
model = HfApiModel(
max_tokens=2096,
temperature=0.5,
model_id='Qwen/Qwen2.5-Coder-32B-Instruct',# it is possible that this model may be overloaded
#model_id='https://pflgm2locj2t89co.us-east-1.aws.endpoints.huggingface.cloud',
custom_role_conversions=None,
)
with open("prompts.yaml", 'r') as stream:
prompt_templates = yaml.safe_load(stream)
agent = CodeAgent(
model=model,
tools=[final_answer,search_terms_of_service,extract_tos_content],
max_steps=6,
verbosity_level=1,
grammar=None,
planning_interval=None,
name=None,
description=None,
prompt_templates=prompt_templates
)
GradioUI(agent).launch()