Skip to content
sandialabsPublic

About

On-Store Intelligence Index (OSII) for AI-Ready Data

Topics

Resources

Stars

4 stars

Watchers

0 watching

Forks

Repository files navigation

OSII — local

Keep the evidence. Grow the intelligence.

OSII (the On-Store Intelligence Index) turns documents and datasets into a library you can browse, search, enrich, and ask questions about. The main branch is OSII-local: a standalone application for working with files on your computer or a shared drive mounted on it. It runs without user accounts or a login; OSII-server provides hosted user access.

The source material is the evidence. OSII keeps originals unchanged and saves extracted content and provenance in portable .osii sidecars. Search indexes are rebuildable, processors are replaceable, and model interpretations remain traceable to their sources. Basic extraction and keyword search work without an AI model.

What you can do

  • Discover shared-drive metadata with resumable scans, then filter or sample a bounded batch for Intake.
  • Extract text, add OCR alongside it, and inspect text or mathematical Markdown with source locations.
  • Browse files and collections, search by keywords or document meaning, and ask grounded questions with an optional model connection.
  • Edit summaries, metadata, wikis, and other derived products with revision history; run replaceable analysis tools.
  • Explore related documents with the experimental Topic Explorer.
  • Compare experimental RAG strategies and inspect their evidence in Chat.
  • Extend processing and frontend tools through documented interfaces.

Get started

Install uv, Node.js 22.18+, and make on macOS/Linux. Copy .env.example to .env, set OSII_SOURCE_DIR to an existing folder, and run:

make dev

On Windows, install uv and Node.js and run:

.\scripts\osii.ps1 dev

Open the dashboard, process your files, and explore the library. Toolbox services start automatically and are available in Setup; choose a tool when you want to use it. Keep the terminal open; Ctrl+C stops OSII. For sample files, use make demo (Windows: .\scripts\osii.ps1 demo). Install Tesseract for OCR; configure optional model connections in Setup. The Banyan Toolbox extractor uses its own local Nemotron endpoint configuration. Keep credentials outside Git. Choose an Intake preset for each batch; the startup guide covers provider profiles and configurable processing defaults.

See the startup guide for setup and troubleshooting, architecture for the design, and extension guides for processors and frontend tools. Use the osii-server branch for an authenticated library shared by a team; archive/workstation-deployment-2026-11-03 preserves earlier workstation packaging workflows.

About

On-Store Intelligence Index (OSII) for AI-Ready Data

Topics

Resources

Stars

4 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages