LLM should send a custom user-agent when using -f https://...

[simonw]

I noticed that Wikipedia blocks the default HTTPX user-agent, as explained here: https://foundation.wikimedia.org/wiki/Policy:Wikimedia_Foundation_User-Agent_Policy

For example:

llm -f https://en.wikipedia.org/wiki/1988_World_Snooker_Championship 'top 3 winners'

Returns:

httpx.HTTPStatusError: Client error '403 Forbidden' for url 'https://en.wikipedia.org/wiki/1988_World_Snooker_Championship' For more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/403

Setting a custom user-agent similar to this should fix it:

User-Agent: CoolBot/0.0 (https://example.org/coolbot/; coolbot@example.org) generic-library/0.0

[simonw]

I'm going to go with:

llm/0.27.1 (https://llm.datasette.io)

[simonw]

Claude Code: https://gistpreview.github.io/?1ade9b28e4698d6bdbf1240e9563fdf4

[simonw]

Testing it:

uv run llm -f https://en.wikipedia.org/wiki/1988_World_Snooker_Championship 'top 3 winners'

Do you mean the top finishers in the 1988 tournament? If so:

• 1st (Champion): Steve Davis — beat Terry Griffiths 18–11 in the final
• 2nd (Runner‑up): Terry Griffiths
• Joint 3rd (losing semi‑finalists): Cliff Thorburn and Jimmy White
  (Semi‑final scores: Davis 16–8 Thorburn; Griffiths 16–11 White)

If you meant something else by "top 3 winners," tell me and I’ll adjust.

uv run llm -f https://httpbin.org/user-agent summary

Summary of the header you provided:

• Raw value: llm/0.27.1 (https://llm.datasette.io/)
• Meaning: an HTTP User-Agent string identifying the client as the "llm" product, version 0.27.1, with a reference URL (https://llm.datasette.io/).
• What it conveys: product name, version, and a homepage/identifier URL. No OS, language, or detailed environment information is included.
• Use: typically used for server logging, analytics, or to apply client-specific behavior on the server.

If you want, I can:

• parse this into structured fields,
• suggest a fuller User-Agent format that includes platform/runtime,
• or generate a regex to match similar user-agent strings. Which would you like?