web-search-and-content

Extract keywords and keyphrases from a text without AI (RAKE algorithm scoring frequency and degree), in Spanish or English, plus most frequent words. For tagging, indexing and SEO.

Search scientific papers on arXiv (AI, physics, maths, economics...): title, authors, abstract, categories, dates and PDF link; by relevance or newest, with category filter (cs.AI, cs.CR, q-fin...).

Search scientific papers on arXiv (AI, physics, maths, economics...): title, authors, abstract, categories, dates and PDF link; by relevance or newest, with category filter (cs.AI, cs.CR, q-fin...).

Extract the tables of a PDF (URL) as JSON rows with their page: financial reports, official lists, invoices... Works on text PDFs (not scans). Optional page range.

AI Document Search Preparation: Split up to 12000 characters of supplied plain text into overlapping chunks and produce MiniLM embeddings. Returns a portable search dataset; not a hosted database. Local processing. Limits and retention: /ai-services.

AI Search Result Reranking: Rank up to 20 passages by relevance to a query using a TinyBERT cross-encoder. Returns original indices and uncalibrated scores, with truncation disclosed. Local processing. Limits and retention: /ai-services.

AI Structured Field Extraction: Extract up to five named fields from up to 1200 characters. Each value must be an exact source substring or null, with missing information left empty. Local processing. Limits and retention: /ai-services.

List items — Structured extraction: Structured extraction for List items. One public HTTPS page or supplied HTML; static parsing, no JavaScript rendering, login or access bypass. Specialised preset in a shared 50-profile, 20-operation engine. 0.01 USDC includes one bounded structured extraction result.

Link labels — Attribute extraction: Attribute extraction for Link labels. One public HTTPS page or supplied HTML; static parsing, no JavaScript rendering, login or access bypass. Specialised preset in a shared 50-profile, 20-operation engine. 0.01 USDC includes one bounded attribute extraction result.

Level-six headings — Structured extraction: Structured extraction for Level-six headings. One public HTTPS page or supplied HTML; static parsing, no JavaScript rendering, login or access bypass. Specialised preset in a shared 50-profile, 20-operation engine. 0.01 USDC includes one bounded structured extraction result.

Code snippets — Attribute extraction: Attribute extraction for Code snippets. One public HTTPS page or supplied HTML; static parsing, no JavaScript rendering, login or access bypass. Specialised preset in a shared 50-profile, 20-operation engine. 0.01 USDC includes one bounded attribute extraction result.

Forms — Structured extraction: Structured extraction for Forms. One public HTTPS page or supplied HTML; static parsing, no JavaScript rendering, login or access bypass. Specialised preset in a shared 50-profile, 20-operation engine. 0.01 USDC includes one bounded structured extraction result.

Figures — Structured extraction: Structured extraction for Figures. One public HTTPS page or supplied HTML; static parsing, no JavaScript rendering, login or access bypass. Specialised preset in a shared 50-profile, 20-operation engine. 0.01 USDC includes one bounded structured extraction result.

Level-four headings — Joined text extraction: Joined text extraction for Level-four headings. One public HTTPS page or supplied HTML; static parsing, no JavaScript rendering, login or access bypass. Specialised preset in a shared 50-profile, 20-operation engine. 0.01 USDC includes one bounded joined text extraction result.

Preformatted text — Joined text extraction: Joined text extraction for Preformatted text. One public HTTPS page or supplied HTML; static parsing, no JavaScript rendering, login or access bypass. Specialised preset in a shared 50-profile, 20-operation engine. 0.01 USDC includes one bounded joined text extraction result.

Document Retrieval Evidence Pack: Compare semantic embedding similarity with cross-encoder relevance, then combine their ranks for up to 5 passages. Returns source indices, both scores and a reproducible evidence report. 0.01 USDC includes up to 5 passages.

Level-one headings — Attribute extraction: Attribute extraction for Level-one headings. One public HTTPS page or supplied HTML; static parsing, no JavaScript rendering, login or access bypass. Specialised preset in a shared 50-profile, 20-operation engine. 0.01 USDC includes one bounded attribute extraction result.

Level-six headings — Joined text extraction: Joined text extraction for Level-six headings. One public HTTPS page or supplied HTML; static parsing, no JavaScript rendering, login or access bypass. Specialised preset in a shared 50-profile, 20-operation engine. 0.01 USDC includes one bounded joined text extraction result.

Tables — Joined text extraction: Joined text extraction for Tables. One public HTTPS page or supplied HTML; static parsing, no JavaScript rendering, login or access bypass. Specialised preset in a shared 50-profile, 20-operation engine. 0.01 USDC includes one bounded joined text extraction result.

Sections — Structured extraction: Structured extraction for Sections. One public HTTPS page or supplied HTML; static parsing, no JavaScript rendering, login or access bypass. Specialised preset in a shared 50-profile, 20-operation engine. 0.01 USDC includes one bounded structured extraction result.

multimodal machine learning: Publication sample breakdown: Publication sample breakdown on multimodal machine learning using Crossref publisher metadata. Returns DOI references and source attribution. Query can be overridden; sample summaries are labelled.

Convert a public webpage URL to clean Markdown for RAG, grounding, research, summarization and LLM context. Returns agent-ready Markdown plus page metadata while stripping navigation, scripts and boilerplate from static public HTML.

Summarize a web page: give a URL, get a short summary of the page written by Claude, with its title and final URL.

Juniper Agent Web Intelligence: one Brave web search, fetch up to three public HTTPS pages and extract source-backed JSON (title, metadata, headings, text, links, JSON-LD). 0.01 USDC on Base per request; no signup. HTML/text/JSON only; no JavaScript rendering or arbitrary LLM schema. Reuse your private request_id on pending responses.