Context Capture Saver

Capture useful context from websites, PDFs, and YouTube transcripts into local JSON, Markdown, and/or SQLite storage you control.

Context Capture Saver is designed for research and AI workflows where you want structured, provenance-aware data instead of ad hoc copy/paste.

Key Features

- Capture article-focused page content, not raw HTML.
- Save with notes, highlights, and queued highlight workflows.
- YouTube transcript capture with transcript-only, timestamped-segments-only, or both.
- Bubble menu support on YouTube selections, including transcript save actions.
- PDF extraction with metadata and selection-aware capture support.
- Local storage backends: `JSON` (default), `Markdown`, `SQLite`, and `JSON + SQLite` (write both).
- JSON output is always written under `json/` inside the selected folder.
- Optional structured JSON chunks (`content.chunks`) in addition to full `documentText`.
- SQLite backend uses a graph-ready schema for analytics and AI agent workflows.
- Automatic migration of legacy SQLite databases (`captures`-only schema) to the latest schema.
- Pending highlights/notes panel now shows queued annotations on-page with delete controls before save.

Capture Workflows

Websites

Use right-click menus, shortcuts, popup, or bubble actions:
- Save content
- Save content with note
- Save content with highlight
- Highlight
- Highlight with note

YouTube

On supported YouTube video pages:
- Save transcript
- Save transcript with note
- Selection bubble includes YouTube transcript actions

Transcript storage modes:
- plain transcript text
- timestamped transcript segments
- both

PDFs
- Extracts text + metadata
- Supports capture and highlight actions
- Browser PDF viewer behavior can still affect some selection UX

Data Storage Overview

JSON Mode
- Default mode.
- Writes one capture file per save.
- All JSON files are written under `json/` in the selected folder.
- Optional setting can include `content.chunks` for agent-ready chunked retrieval while keeping full text.

Markdown Mode
- Writes one `.md` file per capture with YAML frontmatter metadata.
- Body contains the core document text (no structured chunks or timestamped segments).
- Highlights and notes use footnote-style anchors (`[^hl-N]`) placed inline where the highlighted text appears, with a "Highlights & Notes" footer section.
- Markdown files follow the same folder organization options as JSON (date/type grouping).
- Markdown is a lossy format compared to JSON/SQLite -- optimized for human readability, not round-trip data interchange.

SQLite Mode
- Writes/updates a single `context-captures.sqlite` file in the selected folder.
- Uses SQL.js (no external DB server).
- Automatically upgrades legacy DBs in place.

JSON + SQLite Mode

- Writes both outputs in the same save operation.
- If one backend fails and the other succeeds, successful output is preserved and surfaced.

Privacy
- Data is written only to your selected local folder.
- No required remote backend.
- No cloud dependency in capture flow.

Open Source
The plugin is open source and you can see the source code here: https://github.com/stigoleg/Context-Capture-Saver