chat-shiny-python / utils.py
fergos80's picture
Rename utils to utils.py
47070da verified
raw
history blame
3.25 kB
import aiohttp
from bs4 import BeautifulSoup
recipe_prompt = """
You are RecipeExtractorGPT.
Your goal is to extract recipe content from text and return a JSON representation of the useful information.
The JSON should be structured like this:
```
{
"title": "Scrambled eggs",
"ingredients": {
"eggs": "2",
"butter": "1 tbsp",
"milk": "1 tbsp",
"salt": "1 pinch"
},
"directions": [
"Beat eggs, milk, and salt together in a bowl until thoroughly combined.",
"Heat butter in a large skillet over medium-high heat. Pour egg mixture into the hot skillet; cook and stir until eggs are set, 3 to 5 minutes."
],
"servings": 2,
"prep_time": 5,
"cook_time": 5,
"total_time": 10,
"tags": [
"breakfast",
"eggs",
"scrambled"
],
"source": "https://recipes.com/scrambled-eggs/",
}
```
The user will provide text content from a web page.
It is not very well structured, but the recipe is in there.
Please look carefully for the useful information about the recipe.
IMPORTANT: Return the result as JSON in a Markdown code block surrounded with three backticks!
"""
async def scrape_page_with_url(url: str, max_length: int = 14000) -> str:
"""
Given a URL, scrapes the web page and return the contents. This also adds adds the
URL to the beginning of the text.
Parameters
----------
url:
The URL to scrape
max_length:
Max length of recipe text to process. This is to prevent the model from running
out of tokens. 14000 bytes translates to approximately 3200 tokens.
"""
contents = await scrape_page(url)
# Trim the string so that the prompt and reply will fit in the token limit.. It
# would be better to trim by tokens, but that requires using the tiktoken package,
# which can be very slow to load when running on containerized servers, because it
# needs to download the model from the internet each time the container starts.
contents = contents[:max_length]
return f"From: {url}\n\n" + contents
async def scrape_page(url: str) -> str:
# Asynchronously send an HTTP request to the URL.
async with aiohttp.ClientSession() as session:
async with session.get(url) as response:
if response.status != 200:
raise aiohttp.ClientError(f"An error occurred: {response.status}")
html = await response.text()
# Parse the HTML content using BeautifulSoup
soup = BeautifulSoup(html, "html.parser")
# Remove script and style elements
for script in soup(["script", "style"]):
script.decompose()
# List of element IDs or class names to remove
elements_to_remove = [
"header",
"footer",
"sidebar",
"nav",
"menu",
"ad",
"advertisement",
"cookie-banner",
"popup",
"social",
"breadcrumb",
"pagination",
"comment",
"comments",
]
# Remove unwanted elements by ID or class name
for element in elements_to_remove:
for e in soup.find_all(id=element) + soup.find_all(class_=element):
e.decompose()
# Extract text from the remaining HTML tags
text = " ".join(soup.stripped_strings)
return text