link/vault

A visual archive for saved links — not a knowledge base.

Open the bookmark archive →

link-vault turns captured web content into a fast, static bookmark site. Save a URL, keep a readable local copy, and browse the collection by source, date, or search.

The bookmark experience

• visual masonry grid with captured images and readable text fallbacks
• instant search across titles, summaries, authors, and domains
• source filters for X, web pages, Bilibili, YouTube, and WeChat
• grid and compact list views
• local detail pages for every saved item
• responsive layout with no runtime backend

Save a link

Use xfetch as the canonical capture and publish pipeline:

python3 -m xfetch save "https://example.com/article" \
  --content-root ./content-out \
  --json

With a publish target configured:

export XFETCH_TARGET_REPO=/path/to/link-vault
export XFETCH_REPO_OWNER=guchengwei
export XFETCH_REPO_NAME=link-vault

python3 -m xfetch save "https://example.com/article" \
  --content-root ./content-out \
  --json

Publishing flow

URL → xfetch capture → content bundle → Link Vault renderer → GitHub Pages

xfetch owns capture, bundle creation, and sync. This repository owns the public archive renderer and the saved content it publishes. GitHub Pages serves the generated site/ output.

Each bundle contains:

• document.json — source, author, timestamps, publish metadata, and optional card fields
• index.md — the readable captured content
• images and other captured media when available

Better bookmark cards

The renderer accepts optional presentation metadata without requiring it:

{
  "card": {
    "title": "A concise, durable title",
    "opening": "One useful sentence that explains why this link matters.",
    "image": "images/cover.jpg",
    "diagram": "images/diagram.png"
  }
}

The card can use a generated front image, a useful diagram, or a strong opening phrase. Existing bundles continue to work: Link Vault falls back to captured images and extracted text.

Local tools

The repository still supports local CLI bookkeeping and semantic search when useful, but those are secondary to the bookmark site:

python -m linkvault ingest https://example.com/article
python -m linkvault search "machine learning"
python -m linkvault list
python -m linkvault stats

Build the static archive locally:

python -m linkvault.site_builder
python -m http.server 8000 --directory site

Then open http://localhost:8000.

Environment

xfetch must be available on PATH, or configured with XFETCH_CMD or LINKVAULT_XFETCH_CMD.

Publishing uses:

• XFETCH_TARGET_REPO
• XFETCH_REPO_OWNER
• XFETCH_REPO_NAME

Optional defaults:

• XFETCH_BRANCH
• XFETCH_CONTENT_SUBDIR
• XFETCH_SITE_SUBDIR

Live site

https://guchengwei.github.io/link-vault/