🌀 17711 Smart Services

Extracts structured JSON from messy text and, optionally, validates and auto-corrects it against a caller-supplied JSON Schema -- 100% local and deterministic, no LLM call, no network dependency of any kind.

seen 57s ago

Semantic/keyword search over every registered instance's own live tool catalog, so an agent can find which service/tool handles a given task without already knowing the fleet's layout. Split out of service-registry so this service's own local ML embedding dependency stays isolated from the core instance registry.

seen 6s ago

Given a piece of text, suggests 1 to 5 emoji matching its lexicon -- classical lexical/ stemming matching (no ML model, no network call) against a curated multilingual emoji keyword dataset. Works best in English and French (the two languages this dataset was curated for first); best-effort support for 16 other languages. No language field on the request -- the input is stemmed and matched in every supported language, the best-scoring one wins.

seen 44s ago

Checks whether an IP address appears on any of several public IP blocklists (FireHOL level1/level2, Spamhaus DROP, Tor exit nodes, Feodo Tracker, Talos Intelligence): periodically re-fetches each source and merges them into one queryable index -- never a live per-call lookup against the upstream lists themselves. A source that fails to fetch never blocks or empties the others; its own last-known-good entries stay queryable, and its failure is visible via the stats tool rather than hidden.

seen 3s ago

Trivial walking-skeleton service: proves the Rust runtime <-> service wiring end to end with a single echo tool.

seen 27s ago

Maintains a self-verified pool of free HTTP/SOCKS proxies sourced from a public list: periodically re-fetches the source list and, in rotating batches, independently checks each proxy's own liveness/latency by making a real HTTP request through it -- never trusts the source's own claimed alive/latency fields. Read-only tools let other services query the verified pool by country/protocol/anonymity/latency.

seen 13s ago

Generic web scraping: fetches an arbitrary caller-supplied URL, always through a proxy obtained from proxy-pool -- never a direct/unproxied fetch. Retries only on a transport failure (a real HTTP response from the target, whatever its status code, is returned as-is -- never masked or retried away).

seen 6s ago

Internet search, always fetched through a paid proxy obtained from proxy-pool -- every search genuinely consumes and pays for a real proxy, no direct/unproxied fallback.

seen 27s ago