All projects matching “text”

Takes two documents, by URL, inline base64 or raw text, extracts the text of each and compares them. Paragraphs are the unit: the answer lists every block that was removed, added, or kept but reworded, and for reworded blocks it names the individual words that went and came, so a single changed number in a long clause is visible instead of the whole clause being flagged. Alongside the content diff

Free tier: 25 calls/day per IP with header X-Free-Call: 1 (or ?free=1), then pay per call. Rule-based prompt-injection screen for untrusted text such as scraped web pages, emails and tool output. JSON body {text} (max 50,000 chars). Detects instruction-override phrasing, role-spoofing tags, data-leak links and hidden Unicode in 6 languages. Returns verdict safe, suspicious or malicious, a 0-100 sc

Batch text-to-speech: synthesize up to 20 separate text items in a single call, standard voices only, up to 500 total characters, returns JSON with base64 WAV audio plus duration and sample rate per item. Use it to: voice a list of short UI strings at once, generate audio for multiple notification templates, batch-produce voice lines for a menu or form, synthesize several short replies in one requ

Use to extract the main article or document text from a single-page URL. Not for homepages or hubs (Web Index Cleaner) and not a link inventory (Link Extractor). Server HTML only; JavaScript is not run. Output is text or markdown. If the requested extract has no main content, the other profile is tried once in the same request. The HTTP body has correlation_id and data. The item array is data.resu

Repair malformed JSON text into valid parsed JSON: markdown code fences, single quotes, smart quotes, unquoted keys, trailing commas, comments, Python literals (True/None), raw newlines in strings, and truncated output (auto-closes strings, arrays and objects). Returns the parsed value plus a log of every repair applied. Built for LLM output that almost parses. POST a JSON body like: {"text":"```j

YouTube video transcript extraction with per-segment timing. Returns the caption track as timed segments (start, duration, text) plus the full plain text, or as SRT subtitles. Uses the creator's manual captions when they exist and falls back to YouTube automatic captions. Never downloads media, so it stays fast and light. Cached per video and language. Use it to summarize a talk, quote a video wit

Drop-in Exa-compatible search: the same request (query, numResults, category, includeDomains, excludeDomains, startPublishedDate, contents.text, contents.highlights) and response format (results with id, title, url, publishedDate, author, score, text, highlights) as Exa /search, on Google-quality authority-ranked results, $0.006 with page text included (Exa: $0.007 + per-page fees; resellers $0.01

Rule-based multi-label text classification over a fixed taxonomy (urgency, financial_promise, credential_request, crypto_mention, cta, legal_offer, medical_claim, seo_spam, url_shortener, press_release, question, has_contact). Deterministic - identical text always yields identical labels.

Finds personally identifiable information (names, addresses, emails, phone numbers, credit card and bank numbers, SSNs, passwords, IP addresses and more) in English or Spanish text with Amazon Comprehend. Returns each entity's type, offsets and score, and the text with each one replaced by [TYPE].

Answer questions about an image or describe what it shows (Llama 3.2 11B vision), fixed price $0.008 USDC. Use this when the input is an image; use /chat for text-only questions. POST JSON body {image (required, base64 or data:image/... URI, max 4 MB decoded, PNG/JPEG/GIF/WEBP), prompt (optional question, defaults to describing the image), max_tokens (optional 1-512, default 256)}. Returns {text}.

Turn newline-delimited JSON exports into one JSON array for downstream processing. Parse supplied NDJSON or JSON Lines text with LF or CRLF endings. A final newline is required; empty-line skipping is optional. Rejects malformed JSON and duplicate keys. Bounded text input, not file reading or streaming.

Screen up to 3000 characters of untrusted text for prompt injection, secret exfiltration, deceptive payments and phishing. Jev risk scores, five attack families and an exact suspect passage. French, English and Spanish evaluation; probabilistic screening, not a security guarantee. Text is processed by the Jev provider; live credentials are outside the supported input scope.

Prompt-injection scanner for content your agent is about to read. Returns a risk score (0-1), a level (low/medium/high), findings and the cleaned text wrapped as untrusted data (hidden text, invisible characters, data-exfiltration links, instruction-override phrases in 13 languages). Deterministic. Content is scanned in memory and never stored. Not charged if it fails. Also: /v1/read $0.005, /v1/s

Run a regular expression on a text: every match with its position and capture groups (named too), and the text with a replacement applied. RE2 syntax, which runs in linear time (no lookaround or backreferences).

SWARM Read Page - Retrieve the content of a URL as clean, LLM-ready text (headings with #, lists with -; menus, scripts, styles and markup removed). Use when an agent needs to read a web page, article or docs page without paying model tokens for raw HTML: on 15 public pages we measured, the text was a median 97% smaller than the HTML (about 67,000 fewer tokens per page). Requires `url`; HTML or te

Text to tiny, clean SVG code. Best for icons, logos, lettering and badges. Returns real shapes and editable text, each part its own layer. About 10 s.

Read one public web URL and convert its webpage content to clean Markdown or plain text. Removes HTML boilerplate and returns LLM-ready text for AI agents, RAG context, research and summaries. One URL per paid request.

Hacker News stories: front-page lists (top, new, best, ask, show) or full-text search, each with title, URL, domain, points, comment count, author, time, HN link and text excerpt.

Summarize text. Classify sentiment. Rewrite. Extract keywords. Translate. When to use: cheap agent glue / probe steps for buyer-supplied text. Returns JSON for $0.01 per call in USDC on Base via x402. No API key. Local-fast NLP; no login scrape. Success JSON includes next_tasks hints for same-session follow-ups.

Convert kanji-mixed Japanese (JP) text to hiragana, katakana, or romaji (Hepburn/Nippon/Passport), with a per-token breakdown of surface form, reading, and part-of-speech. Backed by the kuromoji morphological analyzer. Use to generate ruby readings, build search indexes, or preprocess text for speech synthesis.

falcon.sol: PDF to plain text or markdown for agents and RAG pipelines. Send a URL or base64 PDF (max 60 pages, 4 MB); get page-by-page text in reading order, with automatic local OCR for scanned pages. format=markdown for headings, bullets and table rows. Empty result is not charged. $0.0025 USDC.

Translation. Translate text to any language. POST /translate {"text":"...","target":"es"}

Healthcare text de-identification engine (demo endpoint — SYNTHETIC/TEST DATA ONLY, never send real PHI; nothing is stored). Strips names, DOBs, member/claim IDs, phones, emails, addresses, SSNs; dates become a derived interval timeline (service-to-denial days, days-ago) so denial and timely-filing math survives de-identification. Returns scrubbed text + token map for local re-identification. Prod

Classify text an AI agent is about to read for prompt injection: verdict (safe, suspicious, malicious), score 0-100, findings with matched spans; decodes hidden Unicode tags, HTML comments, CSS-hidden text and base64 before scanning.

M2M100 machine translation — one x402 payment on embed.gedx402.com. POST { text, source_lang, target_lang } → { translated_text }; wire settlement USDC per request (see npm run pricing:report) worst-case. Hero: GET /heroes/translate. Competitive discovery alongside embed ping; not in the five-step AOV ladder.

Call before publishing a claim that cites EU law. Checks that the cited instrument exists in our pinned corpus, that the cited article exists within it, and whether your quoted text actually appears in that article. Covers the AI Act, GDPR, NIS2, DORA and the Cyber Resilience Act. Returns the official article text with a EUR-Lex link. Citation verification only: it never gives legal advice or clas

Scores text against every category in all six taxonomies (one embedding, cosine similarity) and returns the top `per` rows per taxonomy. With `taxonomy` and `category` it also returns the target's rank, score, gap to #1, and a phrase bank split into phrases the text already uses and phrases it…

Use when you need image embeddings or text embeddings in one 2048-dimension space for retrieval, RAG, semantic search, clustering, or dedup (Qwen3-VL-Embedding-2B, OpenAI-compatible). Batch up to 256 items per call, served 32 at a time. An x402 call is $0.002 when the request includes images, $0.001 for text-only. USDC on Base. 400 is a bad body, 413 is over 48 MiB, 429 means busy (Retry-After).

Payment-Injection Screen by Agentropolis: send text your agent is about to act on (tool result, web page, email, file). Flags payment-looking instructions: pay/transfer requests, fake licence or refundable fees, swapped wallets, token approvals, seed-phrase asks, hidden text. Returns clean/caution/likely_payment_injection with evidence. Advisory only.

Full text of one section of the FDA Investigations Operations Manual (IOM). Get the id from GET /search?q=<keywords> or GET /chapter/:id. Use this when you need the complete, verbatim text of a specific IOM section for inspection-readiness or regulatory research, not a summary.