Machine Shop

Normalize up to 100 texts with NFKC and whitespace cleanup in one payment. Preserve input order; fail the entire batch on invalid input. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Split text into fixed Unicode-character windows with configurable overlap and exact [start,end) offsets. No token counting or semantic boundaries. Maximum 2000 chunks. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Clean whitespace in scraped text, OCR output or pasted documents before indexing, entity extraction, text comparison or LLM input. Deterministic text cleanup: normalize CRLF/CR to LF, collapse runs of ASCII spaces/tabs within each line, trim each line using Unicode whitespace rules, remove empty lines, preserve nonempty line breaks and letter case. Returns text, chars_before (after newline normalization), and chars_after. This is whitespace cleanup, not OCR, HTML parsing, spelling correction or PII redaction. POST application/json. Limit: 50000 input Unicode characters per request; for comparison, the lengths of a and b are added. Empty strings are accepted. Extra JSON properties are ignored. Errors after payment verification: 413 for the character limit; 422 for missing/non-string fields or malformed JSON, and for URL parsing errors. Unpaid calls return 402 with a base64 JSON PAYMENT-REQUIRED header before input validation. Handler error responses are not settled.

Compare text using deterministic, case-insensitive bag-of-words token-overlap similarity for lexical duplicate screening. Lowercase each input and tokenize with Python Unicode regex \w+ (letters, digits and underscores). Count repeated tokens. Similarity = sum(min(count_a(token),count_b(token))) / max(words_a,words_b,1), rounded to four decimal places. Returns exactly similarity in [0,1], words_a and words_b. Word order and punctuation are ignored; both token-empty inputs score 0, even when equal. Does not measure semantic similarity, Levenshtein edit distance, cosine similarity, or produce a text diff. POST application/json with a and b. Limit: 50000 input Unicode characters per request; for comparison, the lengths of a and b are added. Empty strings are accepted. Extra JSON properties are ignored. Errors after payment verification: 413 for the character limit; 422 for missing/non-string fields or malformed JSON, and for URL parsing errors. Unpaid calls return 402 with a base64 JSON P

Convert comma-separated CSV with a unique header row into JSON rows. All cell values stay strings. Unequal row width returns 422; no type guessing. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Normalize up to 100 documents with NFKC and whitespace cleanup, deduplicate by exact resulting text, return unique text and original duplicate indices. No semantic matching. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Normalize and deduplicate up to 100 URLs using existing normalize-url rules; return unique URLs and duplicate indices. No requests or redirects. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Deterministic regex entity extraction and email extraction from caller-supplied text. Returns exactly emails, urls and money arrays, each sorted and deduplicated by exact case-sensitive string equality, preserving matched case. Email patterns use Python case-insensitive [A-Z], digits and common email punctuation with a letter TLD of length >=2, case-insensitive matching. URL patterns start with lowercase http:// or https:// and end at whitespace, angle brackets or quotes; trailing punctuation can remain. Money patterns match $/€/£ followed by a digit and digits/commas/dots, or a number followed by USD/EUR/GBP, with one optional whitespace separator and case-insensitive currency codes. Money is returned as strings, not validated numeric amounts. Currency-code patterns have no trailing word boundary: 5 USDC can yield the partial match 5 USD; USDC is not recognized as a distinct currency. Does not extract people, organizations, locations, dates or standalone crypto addresses; not semantic

Extract caller-selected RFC6901 JSON pointers. Missing paths return 422. Empty pointer selects the root. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Return sorted unique email, URL, money, ISO-date-shaped and EVM-address-shaped matches. Pattern matching only; dates, checksums and deliverability are unvalidated. Money uses legacy USD/EUR/GBP patterns; 5 USDC can produce a partial 5 USD match. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Extract closed HTML anchors and their text; resolve relative URLs against optional base_url. Destinations are untrusted and never fetched. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Extract title, description, language, canonical URL and named/property meta tags from supplied HTML. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Convert supplied static HTML to text and basic Markdown, extract metadata, links and JSON-LD, then normalize, deduplicate lines, extract patterns and chunk the text into a RAG document. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Extract JSON-LD script blocks as parsed JSON with validity flags. No microdata, inference or schema validation. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Convert static HTML headings, paragraphs, lists and anchors to Markdown. Tables, CSS and JavaScript are unsupported. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Remove scripts, styles, templates and markup from static HTML; return visible text with line breaks. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Parse JSON and emit compact recursively key-sorted JSON. Python numeric precision applies; not cryptographic JSON canonicalization. Duplicate keys use last value. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Apply Unicode NFKC then normalize whitespace and remove blank lines; preserve letter case. NFKC can change character meaning. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Normalize URL strings for link deduplication and stable hashing without network access. Deterministic URL normalization using Python urllib.parse: strip surrounding whitespace, require parsed http/https scheme and nonempty authority, lowercase scheme and the entire authority (including any userinfo), use / for an empty path, remove the final path-parameter component and fragment. Decode/re-encode query pairs with urllib.parse.parse_qsl/urlencode, preserving order, duplicate keys and blank values; remove keys beginning with utm_, fbclid, gclid or yclid case-insensitively. Does not sort queries, resolve dot segments, remove default ports or validate port numbers/hosts; ports remain in the lowercased authority. Not a fetch, redirect resolver, security validator or universal canonical URL equivalence test. Returns exactly url and sha256 (SHA-256 of its UTF-8 bytes). Non-http(s), missing authority and parser errors return 422. POST application/json with url. Limit: 50000 input Unicode chara

Produce a RAG document: NFKC and whitespace normalization, exact repeated-line removal, SHA256, pattern entities and overlapping character chunks. Repeated lines are intentionally removed; no semantic analysis. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Repair code-fenced JSON or JSON-compatible Python literals including single quotes and trailing commas. Unsupported repairs return 422; no invented fields. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Fetch one public static HTML/text page and return basic Markdown and final URL. Static fetch: default ports only; public DNS/IPs only; maximum 3 redirects, 1MB, 15s body deadline plus DNS/connect time. No JavaScript, cookies, auth, paywall bypass, browser or paid upstream. Upstream, safety or parsing errors return 422 without settlement. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Fetch one public static HTML/text page and return cleaned text and final URL. Static fetch: default ports only; public DNS/IPs only; maximum 3 redirects, 1MB, 15s body deadline plus DNS/connect time. No JavaScript, cookies, auth, paywall bypass, browser or paid upstream. Upstream, safety or parsing errors return 422 without settlement. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Fetch one public static HTML/text page and return title, description, canonical URL, language and meta tags. Static fetch: default ports only; public DNS/IPs only; maximum 3 redirects, 1MB, 15s body deadline plus DNS/connect time. No JavaScript, cookies, auth, paywall bypass, browser or paid upstream. Upstream, safety or parsing errors return 422 without settlement. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Fetch one public static page and return text, basic Markdown, metadata, links, JSON-LD plus normalized, line-deduplicated, hashed, entity-extracted and chunked RAG content. Static fetch: default ports only; public DNS/IPs only; maximum 3 redirects, 1MB, 15s body deadline plus DNS/connect time. No JavaScript, cookies, auth, paywall bypass, browser or paid upstream. Upstream, safety or parsing errors return 422 without settlement. Limit: 50000 combined input string characters; batch maximum 100. JSON POST; invalid input returns 422, size limit 413. Unpaid production requests return x402 402 before input validation.

Consumer-specific OpenAPI impact and tests; unsupported => REVIEW. POST baseline/candidate/consumer JSON. Idempotency-Key required; 202: reuse authorization/key/input, never pay again. Full schemas: https://api.pay-forge.online/v2/api-migration-pack/schema

Reconcile an Apify-style product dataset against a previous accepted catalog and explicit flat field contract. Return ingestion-ready inserts/updates, unchanged records, quarantines, duplicate resolution, money/unit checks, schema drift and field source pointers. Upsert only; no inferred deletion, fetching or external purchases. Maximum 50 current and 50 previous records, 16 scalar fields, 100000 input bytes. Fixed 0.050000 USDC per batch.