Back to home

Changelog

What changed in each release of Return.

  1. 1.1.0

    Batch processing for large documents, a smarter model library, and PDF export fixes.

    Added

    • Analyze documents too large for a single pass - Return splits them, works through each part, and combines the results automatically.
    • Imported PDFs and files opened from outside your workspace are now searchable by meaning in chat.
    • The model library now highlights the best local model for your Mac.
    • The file explorer now highlights the file you are working on and expands its folders in the tree so you can see where it lives.
    • Documentation now includes screenshots, and inline UI icons - rendered both in the app and on the website.

    Changed

    • Refreshed the local model catalogue with newer models, each described by what it is good at.
    • The command palette (type "/") now runs a workspace search directly and lists only commands that do something.

    Fixed

    • AI features now always use the model you selected in Settings.
    • More accurate memory estimates for the latest models, so they are no longer flagged as too large for Macs that can run them.
    • Sharper assessment and annotation matching.
    • PDF export now wraps long lines, renders tables, keeps your editor line breaks, and large exports open correctly in Preview and other Apple apps.
    • Assessment highlights and chat history stay attached to your document when you reopen it - and even when you edit the file in another app.

  2. 1.0.0

    First public release of Return.

    Added

    • Local-first document analysis with two bundled llama.cpp engines - fully offline, your documents never leave your Mac.
    • Returns - reusable AI prompt templates with Traffic Light and inline annotation outputs.
    • Semantic search across your workspace, powered by on-device embeddings.
    • Cloud AI (Pro) via an Anthropic-backed proxy.