Add SerpKite as a web search (and optional scraper) provider

[Nolikzero]

Opening this here first because CONTRIBUTING asks for an approved proposal before work starts on a feature that isn't on the roadmap.

Disclosure: I work on SerpKite, so I'm the vendor, not a user reporting a need. I haven't written any LibreChat code for this and won't unless you say it's wanted. "No" or "not now" is a perfectly good answer.

Proposal

Add SerpKite as an opt-in entry in SearchProviders (currently serper, searxng, tavily, keenable) and, optionally, in ScraperProviders. SerpKite is a Google SERP API with a URL-to-Markdown endpoint, so it can fill either role or both, the way Tavily does. Operators who never set serpkiteApiKey would see no change.

API contract

GET https://api.serpkite.com/v1/search?q=<encoded>&num=<1-100>&country=<gl>&language=<hl>&time=<hour|day|week|month|year>
Authorization: Bearer skt_live_...

• results[]: { position, title, link, domain, displayed_link, snippet, date }. Top-level answer_box, knowledge_graph, people_also_ask, related_searches and top_stories map onto the existing serper-shaped fields (organic, answerBox, knowledgeGraph, peopleAlsoAsk, relatedSearches, topStories) without loss.
• GET /v1/news uses the same parameters and returns { title, link, snippet, source, date, image_url }.
• Scraper: POST /v1/webpage with {"url": "<public URL>"} returns the page as Markdown, with an SSRF guard on our side. LibreChat's own allowedAddresses preflight would still apply.
• Errors always have the same shape: {"error":{"code","message","request_id"}}, with 400 (unknown or invalid parameter, since params are strict), 401, 402 (out of credits), 429 and 503. The adapter can surface error.message directly.
• Things to know up front: num is capped at 100. time takes the five named windows, not custom ranges (those go in separate start_date/end_date params). include_domains/exclude_domains are native parameters, so site: rewriting isn't needed. Failed and empty searches aren't billed.

Docs: https://serpkite.com/docs. OpenAPI: https://serpkite.com/openapi/serp-api.yaml.

Affected areas

The closest template is the Keenable work (#15288) and Tavily (#12581):

• packages/data-provider/src/config.ts and config.spec.ts (enum entries, serpkiteApiKey and an optional serpkiteApiUrl)
• packages/data-provider/src/types/web.ts, packages/data-schemas/src/app/web.ts and types
• packages/api/src/web/web.ts (auth and credential resolution) plus specs
• client/src/components/SidePanel/Agents/Search/ApiKeyDialog.tsx and en translation keys
• librechat.example.yaml, .env.example
• The provider adapter in @librechat/agents (search and scrape functions next to the serper and tavily ones), as a separate PR there

Everything is additive. I'd follow the existing SSRF, fail-closed credential and "system key never forwarded to a user-controlled URL" rules from #15288.

What I'd ask

Is a SerpKite provider something you'd accept? If so, should it be search only, or search plus scraper? I can open the @librechat/agents adapter PR first so you can review the mapping on its own. Happy to provide a test key for CI or review.