I just added a bit of code to my site’s intake endpoint to create #LLM generated descriptions from my blog posts so that when I upload a thought from ia writer, it gets a description for the /blog/ index page.
if not description and settings.OPENAI_API_KEY:
openai.api_key = settings.OPENAI_API_KEY
response = openai.chat.completions.create(
messages=[
{
"role": "system",
"content": "You are a helpful assistant that generates short descriptions for blog posts."
},
{
"role": "user",
"content": f"""Generate a short description for the following blog post content:
<content>
{content}
</content>.
For example: 'Quick notes on my interview on the Hard Part Interview podcast.' or 'I made a thing that converts your pocket saves into an rss feed'
Do not write anything else other than the description and do not wrap the description in quotes."""
}
],
model="gpt-4o-mini",
)
description = response.choices[0].message.content
Pretty simple stuff, but it works well with just a couple of past examples. When I didn’t have the examples, the descriptions didn’t sound like my voice, but with them, they’re not bad and cheap as chips.
Later, I’ll add a prompt to create tags, choosing from the ones I’ve already created.