Import TILs as blogmarks

[simonw]

Having a separate https://til.simonwillison.net/ blog for TILs means I have content that isn't represented in my main blog (aside from when I link to them in my weeknotes).

I want to automatically populate a blogmark for each TIL. I can use a similar approach to this one for that: https://github.com/simonw/simonwillisonblog/blob/main/blog/management/commands/import_blog_json.py

I wrote about that here: https://simonwillison.net/2017/Nov/4/import-refs/

[simonw]

It already has the ability to import blogmarks:

simonwillisonblog/blog/management/commands/import_blog_json.py

Lines 72 to 83 in f0caa71

	elif item["type"] == "blogmark":
	klass = Blogmark
	kwargs = dict(
	slug=slug,
	link_url=item["link_url"],
	link_title=item["link_title"],
	via_url=item.get("via_url") or "",
	via_title=item.get("via_title") or "",
	commentary=item["commentary"] or "",
	created=created,
	metadata=item,
	)

And import_ref can be used to avoid importing the same thing twice:

simonwillisonblog/blog/management/commands/import_blog_json.py

Lines 86 to 89 in f0caa71

	if item.get("import_ref"):
	obj, was_created = klass.objects.update_or_create(
	import_ref=item["import_ref"], defaults=kwargs
	)

So... I could teach it to pull a feed from https://til.simonwillison.net/ in response to a POST to some endpoint.

[simonw]

Example JSON: https://gist.github.com/simonw/857572d9b36cd1e791c730790ed489ef

[simonw]

Five of my blogmarks are already to TILs: https://simonwillison.net/dashboard/?sql=select+%2A+from+blog_blogmark+where+link_url+like+%27https%3A%2F%2Ftil.simonwillison.net%25%25%27%3AA14XB8ah9tdq44NaEbKMaLFU3Q7RYbfOm3ZBoEh_iKU - I could update these in the Django admin with the expected import_ref to avoid creating duplicates.

[simonw]

One catch: what should the link description be?

A few options:

• Take the first paragraph of the TIL, with tags stripped - like on https://til.simonwillison.net/datasette
• Go back through all 324 existing TILs and add a short summary to each one
• The fun one: get GPT-3 to create summaries for everything and then manually write summaries going forward