Web Page to Markdown, JSON-LD & RAG Chunks (LLM-ready document)
Pricing
$4.00 / 1,000 documents
Web Page to Markdown, JSON-LD & RAG Chunks (LLM-ready document)
Convert any web page or docs site into clean, LLM-ready data: HTML to markdown, parsed JSON-LD and OpenGraph structured data, headings, tables, the link graph, token counts and embedding-ready chunks. Built for RAG pipelines, AI agents and content analysis. No browser, no proxies, pay per document.
Pricing
$4.00 / 1,000 documents
Rating
0.0
(0)
Developer
Andy Wang
Maintained by CommunityActor stats
0
Bookmarked
2
Total users
1
Monthly active users
9 hours ago
Last modified
Categories
Share
Turn any web page or docs site into one clean, structured document for RAG, agents and analysis: HTML to markdown, structured data, and embedding chunks in a single dataset item per page. You get markdown, structured data, the link graph, token counts and ready-to-embed chunks — in a single dataset item, per page.
Most "web to markdown" tools hand you a wall of text. This one keeps the shape:
markdown— article content with navigation, footers, cookie banners, ads, scripts and styles removedjsonLd— everyapplication/ld+jsonblock, parsed (Article, Product, FAQ, Recipe, Organization …), broken blocks skipped instead of failing the runopenGraph—og:*metadata as an objectheadings— h1–h3 with levels, so you can cut sections without guessingtables— every table as an array of rows, for data the text loseslinks— resolved absolute URLs with anchor text, plus internal/external countstokenEstimateandwordCount— so you can budget a prompt before you send itchunks— overlapping, paragraph-aligned chunks sized for embedding
Input
| Field | Default | What it does |
|---|---|---|
startUrls | — | Pages to convert; one dataset item each |
maxPages | 1 | Hard cap per run (1–50) |
followSameHostLinks | false | Queue same-host links (docs site → one dataset) |
includeChunks | true | Emit embedding-ready chunks |
chunkTokens | 800 | Approximate tokens per chunk |
timeoutSecs | 30 | Per-page timeout |
Output example
Pricing
Pay per document. No subscription, no minimum. A single page costs a fraction of a cent; a 30-page docs crawl is priced per page.
Notes and limits
- Fetches the HTML directly: fast and cheap, but it does not execute JavaScript. Pages that render client-side need a browser actor — say so in a review and we will add an optional rendering mode.
- Respects
robots.txtfor followed links, caps responses at 5 MB, and refuses non-HTML content types rather than returning garbage. - Nothing is stored beyond your run's dataset.
Built and maintained by the team behind deepsense.stream. Issues and feature requests: open an issue on the actor page — every review is read.
