Open File Bridge

Give Open WebUI controlled access to a user-selected local folder, with local-only storage and an allowed-site lock.

Turn Open WebUI into your Cowork — one that works in ONE folder on your own computer: the folder you pick. Your AI chat can then list, read, write, search, OCR, and convert files in that folder, while everything stays on your machine. Why: Open WebUI's code interpreter runs in a sandbox that cannot touch your files. Open File Bridge gives it a locked door — you choose the folder, you choose the ONE website allowed to use the door, and the extension enforces both. How it works: 1. Pick a folder in the extension's settings. 2. Allow your Open WebUI address (for example http://127.0.0.1:8080 or https://chat.example.com) and optionally set a bridge token. 3. Install the bridge skill in Open WebUI (copy-paste from the GitHub repo) and turn on the code interpreter in a new chat. 4. Ask: "list the files in my folder", or "write the summary to notes.md". Safety model: - One folder only — nothing outside it is visible. - One site only — every other origin is refused; an optional bridge token stops a fake page impersonating your Open WebUI address. - Approve/Deny cards inside the chat page for deletes and overwrites; every overwrite keeps the previous version; deletes go to a 30-day trash. - Read-only mode, write-rate brake, ignore patterns, and built-in refusal of credential-looking files (.env, id_rsa, *.pem, …). - Local-only: no cloud, no accounts, no analytics, no telemetry. Files go nowhere except to your own Open WebUI when it asks. Extras: OCR in 20+ languages (runs locally in the extension), PDF text extraction, xlsx→csv and docx→html conversion, and full-text search. Where your Open WebUI must run: anywhere you can install the bridge skill — on the same computer (localhost), on your LAN, or on a server you control. For a public HTTPS deployment, also set a bridge token. Hosted Open WebUI services that don't let you install skills aren't supported. Requires Open WebUI with the code interpreter enabled. The bridge skill file, setup guide, and source code: https://github.com/Sparkling-AI/open-file-bridge