TinySearch v0.5.0

TinySearch 0.5.0 makes web retrieval simpler and more useful for MCP agents.

Highlights

Let agents formulate retrieval queries. Agents can now rewrite, clarify,
translate, narrow, spell-correct, and add relevant temporal context before
calling research or scraping a page. Tool guidance keeps important names,
constraints, qualifiers, negations, and user intent intact.
Fast discovery is now the default. Use search(query) for lightweight,
backend-ordered web discovery, with XML-formatted results that are easy for
agents to consume.
A simpler scraping surface. scrape_url has been replaced by
scrape_urls(items), which handles one to five independent URL/query pairs
and returns per-item outcomes. Omit an item's query, or use *, for clean
Markdown in page order; supply a focused query only when chunk selection is
useful.
Clearer time awareness. get_current_datetime() now returns XML, so
agents can orient time-sensitive searches before retrieving information.

Replace scrape_url(url, query) calls with
scrape_urls([{"url": url, "query": query}]).
Move agent query planning into the agent: retrieval queries do not need to
repeat the user's wording verbatim.
research(query) remains available but is deprecated; prefer search(query)
for discovery.

Full Changelog: v0.4.3...v0.5.0