USER_MANUAL.md

PageParse User Manual

Welcome to PageParse, a secure, local, offline-first application designed to parse and extract structured task logs, notes, and details from images, text documents, voice memos, spreadsheets, and video files.

Features

• Multimodal Extraction: Supports images, voice files, video recordings, spreadsheets, and scanned documents.
• Local AI Explanations: Leverages Ollama vision models (moondream, llava) and SLMs (llama3.2) to generate detailed explanations and structure raw layout fields locally.
• Structured SQLite Logging: Extracted items are classified and parsed into a lightweight SQLite database.
• Air-gap Fallback: A visual mode switch allowing zero-network mock storage capability.

---

Operating Instructions

1. Prerequisites

Ensure Ollama is running locally with the necessary models:

ollama run llama3.2:1b
ollama run moondream

2. Start the Backend

Activate your virtual environment and start the FastAPI web server:

.venv\Scripts\activate
python -m pageparse.cli serve

3. Accessing the Web App

Open your browser and navigate to: http://localhost:8000/

4. Uploading Media

1. Ensure the Air-gap mode switch is toggled OFF in the top-right settings.
2. Select your media file (image, audio, or video) inside the dropzone.
3. Wait for progress to complete (100%).
4. Click on the item in Ingestion History to see the detailed local AI summary in the Raw Extract panel.