Add Live Tennis API integration

[bensynapse]

Adds the Live Tennis API integration page — live tennis scores, matches and players as Haystack Documents for RAG and agent pipelines (ATP, WTA, Challenger, ITF).

• Package: livetennisapi-haystack on PyPI (published, trusted-publishing/OIDC)
• Repo: https://github.com/livetennisapi/livetennisapi-haystack — 31 tests, MIT
• Two components: LiveTennisMatchFetcher (live/upcoming/completed matches → Documents) and LiveTennisPlayerSearch; both follow the Haystack secret-management conventions (Secret.from_env_var)
• Users bring their own free API key (1,000 req/day, no card): https://livetennisapi.com/subscribe/free

[vercel]

Someone is attempting to deploy a commit to the deepset Team on Vercel.

A member of the Team first needs to authorize it.

[kacperlukawski]

Thank you, @bensynapse Could you please help me understand if that package was created based on an actual need of a user? I would like to determine if that is something useful for a broader audience.

[bensynapse]

Thanks for asking directly, @kacperlukawski — honest answer:

No, it did not come from an inbound Haystack user request. I'm the developer of the Live Tennis API itself, and this is a vendor-authored integration: I built it so Haystack pipelines can pull live tennis data (scores, players, fixtures, win-probability) without hand-rolling an HTTP component, and I submitted it as part of making the API usable across the agent/RAG ecosystem (we ship the same capability as an MCP server and integrations for several other frameworks).

So the demand signal today is "the API has users; none of them have specifically asked for Haystack" — I'd rather say that plainly than imply otherwise. The concrete audience I had in mind is sports-media and betting-analytics teams building QA/agent pipelines over live sports data, where the fetcher plus a well-typed document shape is the fiddly part this package does once.

If the bar for the integrations page is demonstrated end-user need, I completely understand holding this until someone asks for it — happy to leave the package published on PyPI and revisit, or adjust scope if a leaner example-based listing fits better.