Skip to content

PRODUCT.md

Latest commit

 

History

History
35 lines (29 loc) · 2.09 KB

PRODUCT.md

File metadata and controls

35 lines (29 loc) · 2.09 KB
Raw

ParseKit – Product Brief

Overview

ParseKit is a lightweight, native-feeling macOS menu-bar app that batch-converts folders of PDFs, Office docs, and images into clean text, Markdown, or JSON — ready to drag into ChatGPT, Claude, Grok, local RAG, or any knowledge tool. It uses LiteParse v2 (Rust-native) under the hood but hides all complexity.

One-Line Description

ParseKit turns a folder of documents into text files you can paste anywhere — no Terminal, no cloud upload.

Target Users

Knowledge workers, researchers, lawyers, consultants, and power users who live in PDFs/Office docs and want instant, private, local preprocessing before feeding LLMs or building RAG indexes.

Core Jobs-to-be-Done

  1. Quick ingest: folder of PDFs → clean Markdown/text files in seconds.
  2. Routine batch processing: normalize incoming regulatory docs/contracts into a consistent output folder.
  3. Pre-processing for vector DBs & local RAG: produce predictable, layout-aware output.

Functional Requirements

  • Menu-bar icon → elegant popover panel (Tauri tray + custom HTML window).
  • Controls (all persisted):
    • Input folder picker (recursive).
    • Output folder picker.
    • Output format selector: Text (.txt), Markdown (.md), JSON (.json).
    • OCR toggle (default ON) + language selector (default "eng").
  • One-click "Run Parse" button.
  • Live progress: global bar + per-file list (Parsed / Skipped / Error).
  • Recent batches (last 5) with one-click "Open in Finder".
  • Post-batch: "Open Output Folder", "Copy last Markdown to clipboard".
  • Smart Markdown output: # Title + ## Page N markers + --- separators.

Non-Functional Requirements

  • 100% local (LiteParse v2 native OCR; optional first-run Tesseract language data).
  • Native macOS look & feel (dark mode, SF Symbols, accessibility, haptics).
  • Zero telemetry. Network use is limited to optional GitHub release checks (updater) and first-run OCR language data downloads when needed.
  • Minimal CPU/memory; smart concurrency (max 4-6 workers on M-series).
  • Ad-hoc signed .app builds (see docs/RELEASING.md); notarization is planned for public distribution.