suchi

Frequently asked questions

Does it work offline?

The server runs on your own machine and makes no outbound calls until an operator enables an integration such as mail, OIDC, or a hosted model. Suchi Companion can keep selected documents as full offline copies. New captures upload only while the app is in the foreground and the server is reachable; failed items remain visible for retry. See the Companion guide.

Where is my data stored?

Persistent state lives under one data directory: SQLite for metadata, search, identity, and permissions; original and derived files in content-addressed blob storage; and a human-readable but rebuildable rendered/ filing tree. An exact recovery needs a consistent backup of the complete data directory, not only the database or rendered tree. See backup and restore.

Does it send my documents to an AI service?

Not unless an operator configures one. A local model through Ollama or llama.cpp keeps model traffic local. Pointing Suchi at a hosted OpenAI-compatible service requires explicit acknowledgement that document text will leave the server. Mail, OIDC, and other integrations also create only the outbound paths the operator enables.

Can I move from Paperless-ngx?

Yes. suchi import --from <export dir> reads a compatible Paperless-ngx export. Archive serial numbers are retained when valid and available, and values that cannot be carried across are listed in the import report. Test the import against a copy and keep the source export until the result has been checked.

Does it support ASN labels?

Yes. Print a payload such as ASN4021 as a QR code with the decimal number visible beside it on the first page. assign mode keeps one PDF intact and records its number; split mode turns one feeder scan into one document per retained label and rejects an ambiguous batch before importing any child. Read supported formats and configuration.

Does it work with non-English documents?

Search finds terms inside German compounds and in Chinese, Japanese, Korean, and Thai text. Suggested dates recognise month names or number forms across 17 languages and several scripts, and typed monetary values accept common local formats. OCR quality still depends on the installed Tesseract language data and the source image.

What do I need to run it?

Docker, or one Suchi binary plus Tesseract for OCR. Additional converters are optional and enable the formats that need them. The server itself needs no PostgreSQL server, Redis instance, or queue service.

Is it free?

Yes. Suchi is available under AGPL-3.0 with no paid tier. The source is published in the main repository.

Is it stable?

Suchi is an early release. Back up the complete data directory, test a restore before relying on the archive, and read the documented database compatibility policy before upgrading. Keep ordinary-file takeouts as a separate portability path; they are not exact instance restores.