Web & Data

Convierte un texto en un vector (embedding) para búsqueda semántica, RAG, clustering o memoria. El bloque base que todo agente de IA necesita. Alto volumen. input=texto.

Devuelve la respuesta de un LLM (Kimi/DeepSeek) a tu consulta. Úsalo para inferencia de propósito general pagada por llamada cuando un agente necesita una respuesta de modelo. input=tu consulta en texto.

Detecta el idioma de un texto (código ISO). Barato y de alto volumen para agentes. input=texto.

Descarga una página web y devuelve su texto limpio y legible (sin HTML ni scripts). Úsalo para que un agente lea el contenido de una URL pagando por llamada. input=URL http(s).

Clasifica un texto por tema/categoría. Alto volumen para enrutar/etiquetar. input=texto.

Repara/valida JSON roto y lo devuelve limpio. Herramienta dev de alto volumen. input=JSON o texto.

Rankea una lista de documentos por su cercanía de SIGNIFICADO a una consulta (similitud coseno sobre embeddings). RAG listo para usar, sin que el agente monte su propia base vectorial. input='consulta || documento 1 || documento 2 || ...' (separados por ||).

Traduce una instrucción en lenguaje natural a una estructura/contrato densa y accionable. input=texto.

OCR: extrae el texto de una imagen (JPG/PNG) o de un PDF, incluidos PDFs escaneados sin capa de texto. Motor local (Tesseract 5), idiomas español e inglés. Sube el archivo por multipart (campo 'archivo') o mándalo como 'archivo_b64' en el JSON. Opcional 'lang' (spa|eng|spa+eng).

Fetch any URL through a real residential/mobile device (UpRock's network, 190+ countries) instead of a datacenter IP -- returns the page as an actual user in that location would see it, bypassing anti-bot blocking and geo-filtered content.

Everything a reviewed public page declares, in one call and one fetch: title, meta and Open Graph, canonical, robots directive, JSON-LD with its schema.org types, resolved links with an external count, tables, content fingerprints, and readable Markdown. Replaces six separate extraction calls, and adds the cross-checks none of them can make alone — a canonical that disagrees with the URL you asked for, a page declaring noindex, missing structured data.

Extract one supplied HTML document into machine-readable evidence and cross-check it against itself. Returns title, metadata, canonical, robots directive, JSON-LD, resolved links, tables, readable text and content hashes. Reports a canonical that disagrees with the URL served, an Open Graph title that disagrees with the document title, a noindex directive, and absent title, description or structured data. Scripts are not executed and nothing is fetched.

Web search, page extraction, and multi-hop web research.

Touch the data

Extrae datos estructurados de un texto libre y los devuelve como JSON. Úsalo para parsear texto a datos. input=texto.

Aggregate defense statistics from in-kernel XDP/eBPF across the fleet: blocked packets, event counts, active source ranges and PoI-minted integrity proofs. Kernel counters, not application logs.

A specific dated snapshot (YYYY-MM-DD, ids listed free at /) of the 129-series AI-progress and technology pull: frontier benchmark scores, compute trends, AGI prediction-market odds, model pricing, AI stocks, energy, space, robotics, research output. Same per-source shape as /feed (latest, previous, % change, source URL). Use it to compare dates yourself, or call /delta to get the comparison computed. No composite score. Unknown id -> 404.

Scrape the first page of any public Airbnb search URL (city, dates, guests, filters) into an array of listing cards with id, title, nightly price, currency, rating, review count and cover photo. Up to 18 results per call.

Minute bars with the taker buy versus total quote volume split, pre-joined across venues and kept beyond exchange retention: the raw input behind order flow imbalance and VPIN work, in one query instead of paginating an exchange API per symbol per venue.

Search Polymarket markets by keyword and return matching questions with live outcome prices

Headless-Chrome screenshot of a URL (PNG, base64-encoded).

Fetch any URL and return parsed Open Graph / Twitter Card / standard HTML metadata: title, description, image, favicon, canonical, site name, type, author, keywords, published time.

Pull named fields out of a short blob of text as JSON. Use it for: Extract name, email and company from this signature block.; Extract amount, currency and due date from this invoice line.

See how declared response statuses changed per operation, and which operations are missing the statuses you require. Reports statuses added and removed, operations declaring no 2xx response at all, and responses with no description. Useful when a client has exhaustive error handling that a silently added status will fall through.