pi-tiny-search

extensionmaintained

Instant web search and page fetching for Pi via Monid/TinyFish. No browser, no LLM summarization — just fast structured results.

by — · v0.1.1 · published 5d ago

$ pi install npm:pi-tiny-search
downloads/mo
166
stars
0
last push
5d ago
open issues
0

Signals

license: MITtestspi manifest: missinginstall size: —deps: 0peer deps: 0

Download trend

345 downloads · last 12 weeks (weekly)

README

pi-tiny-search

Instant web search and page fetching for Pi. No browser, no LLM summarization — just fast structured results via Monid/TinyFish.

What it does

Two tools, one command:

ToolWhat it does
tiny_searchSearch the web. Returns title, url, snippet, site_name per result.
tiny_fetchFetch a URL. Returns clean extracted text from the page.

Both are pure API calls — no browser rendering, no LLM summarization step. Results come back instantly.

⚡ Performance

Real API benchmarks. 3 runs, 27 search queries + 20 fetch URLs per run. No mocking. All numbers are median (p50) averages across runs.

Latency at a Glance

SearchFetch
Median1881ms2186ms
p902473ms2926ms
p952790ms3014ms
Best1536ms1687ms
Reliability72/72 (100.0%)—

🚀 Concurrency: Where It Gets Fast

Fire multiple requests in parallel. Effective latency = wall time / count.

Parallel RequestsSearch Effective LatencyFetch Effective Latency
1 (sequential)1881ms2186ms
3749ms (2.5× faster)892ms (2.5× faster)
5436ms (4.3× faster)585ms (3.7× faster)
8281ms (6.7× faster)—

8 parallel searches cost the same wall time as 1. At 35ms effective latency per query, the agent gets answers faster than it can process them.

Burst Stability

10 sequential searches back-to-back — no throttling, no errors:

MetricValue
Success rate10/10
Avg latency1933ms
Min–Max1647ms – 2458ms

What This Means

Use CaseLatencyNotes
Single search~1.9sFaster than opening a browser
Single fetch~2.2sClean text, no rendering
8 parallel searches~281ms totalSame time as 1 search
Burst (10 queries)~1.9s eachNo rate limiting

Benchmarked 2026-09-19 with 3 full runs across 27 search queries (short, medium, long, filtered, edge cases) and 20 fetch URLs (fast, slow, multi-URL, selectors, formats). All tests hit the live Monid/TinyFish API — no mocks.

Setup

pi install npm:pi-tiny-search

Then run /tiny-search login in Pi to configure your Monid API key. Or set TINYSEARCH_API_KEY in your environment.

Usage

The agent can call the tools directly:

tiny_search(query: "TypeScript 5.8 release notes")
tiny_fetch(url: "https://devblogs.microsoft.com/typescript/")

Or use the command for setup:

/tiny-search login     # Store your Monid API key
/tiny-search status    # Check auth and config state
/tiny-search enable    # Enable for this session
/tiny-search disable   # Disable for this session
/tiny-search logout    # Remove stored key

Configuration

See docs/configuration.md for the full reference.

License

MIT