toolsDetail.backToAllTools
toolsDetail.projectRecordtoolsDetail.categories.gbptoolsDetail.statuses.livetoolsDetail.updatedOn

OKF Builder (Remix)

Crawl a business website, validate the result, explore its knowledge graph, and export a ready-to-host OKF v0.1 bundle.

OKF Builder (Remix) turns a public business website into a structured Open Knowledge Format v0.1 bundle. It combines a multi-page crawler and Gemini extraction with a visual graph, raw Markdown and YAML editing, conformance checks, browser autosave, and single-click ZIP export.

toolsDetail.openTooltoolsDetail.privateRepository

toolsDetail.overview

Enter a public website URL and the app crawls up to 25 same-origin HTML pages, strips away navigation and presentation markup, and asks Gemini to turn the remaining source text into linked Business, Service, Location, FAQ, and Reference concepts. Generated concepts carry source URLs and citations, and the extraction prompt explicitly tells the model not to invent facts that cannot be tied to the crawl.

The generated bundle remains a working document. You can search its concepts, inspect the relationship graph, edit structured fields, or switch to the raw Markdown and YAML representation. The validator checks required concept types and fields, ID format, timestamps, and type-specific content, while malformed frontmatter is rejected without replacing the current concept.

When the bundle is ready, the browser packages it as a self-contained ZIP with concept Markdown, root and changelog indexes, an llms.txt discovery file, deployment notes, and a standalone D3 graph visualizer. Drafts are autosaved in local browser storage, making the tool practical for reviewing and refining a crawl before publishing the files on a business website.

toolsDetail.keyFeatures

Grounded multi-page extraction

Crawls same-origin pages from a public business website, then uses Gemini to create structured concepts with source URLs, citations, tags, timestamps, and bundle-relative links.

Graph and document workspace

Explore concepts in a D3 relationship graph, search the bundle, inspect individual files, and edit either structured fields or their raw Markdown and YAML.

Conformance checks while editing

Checks required fields, valid concept types and IDs, timestamps, bundle composition, and type-specific content; invalid raw frontmatter is reported without applying the broken edit.

Complete, ready-to-host export

Builds a ZIP containing the concept files, index.md, log.md, README.md, llms.txt, and a standalone viz.html graph explorer entirely in the browser.

Autosaved browser workspace

Keeps the current draft in localStorage and restores it on the next visit, so a review session can continue without sending draft state to a separate persistence service.

Input and prompt-injection safeguards

Sanitizes submitted input, limits crawl scope and page size, parses model output through a response schema, and treats edited content as raw strings rather than executable markup.

toolsDetail.architecture

ApplicationNext.js 15 App Router application with React 19, TypeScript, Tailwind CSS, Motion transitions, and server-side API routes.
Crawler and AIServer routes use Cheerio to extract text from same-origin HTML pages and the Google Gen AI SDK with Gemini structured output to produce OKF concepts.
Validation and editingA TypeScript validation layer parses YAML frontmatter, validates individual concepts and the full bundle, and feeds the React workspace's structured and raw editors.
Visualization and exportD3.js renders the concept graph; JSZip assembles the self-contained bundle and visualization client-side before download.
HostingThe public instance is deployed on Google Cloud Run in asia-southeast1. Draft workspace state is stored in the user's browser.
Next.jsReactTypeScriptGemini APICheerioD3.jsJSZipGoogle Cloud Run

toolsDetail.bottomCtaQuestion

toolsDetail.browseAllTools