Document Processing

Documents can be processed locally in the browser or on our server, depending on settings and task. Server-side processing is optional and helps make large or technically difficult documents usable more reliably.

What happens locally

When a task is processed locally, elmtree chat tries to prepare the content directly in the browser. This minimizes data transfer, but has limits depending on browser, file size, and document type.

Local processing is especially suitable for short texts, simple documents, and confidential content that should not leave your device.

When the server can help

Our server can extract, prepare, or index documents for RAG when local processing is not enough and you choose to use this support. This mainly affects large PDFs, complex documents, OCR-heavy content, and documents that should support repeated follow-up questions.

Server-side processing does not mean that we analyze your documents. We provide optional compute for your concrete task.

What we do not do with document data

We do not analyze uploaded documents. Not for training, not for advertising, not for profiling, and not for product analytics.

Depending on the function, technical metadata can exist, such as file size, status, expiration time, or error state. This metadata helps control processing and deletion; it is not content analysis.

Pseudonymization

Pseudonymization remains a separate protective layer. Local or server-side document processing decides where technical preparation happens. Pseudonymization decides which version later enters the chat context and goes to the model.

More details are available under Pseudonymization and RAG Temporary Storage.