DeepTutor v1.7.2 Release Notes

Release Date: 2026.11.06

Building on v1.7.1, v1.7.2 adds selected-text read-aloud, easier model setup, and more reliable guided learning, retrieval, and updates. Desktop releases now require Apple Silicon and Intel Mac installers alongside both Windows downloads.

What's New

Complete desktop downloads and sturdier Windows support

GitHub Releases provide Apple Silicon and Intel DMGs, a Windows installer, and a Windows portable ZIP. The release workflow checks every platform and all versioned downloads and stable aliases. Windows fixes cover session and migration locks, plugin environments, Python executable discovery, non-UTF-8 output, and paths containing spaces.

Read just the passage you need

Select text in Chat, Mastery Path, or Reading and choose Read aloud to hear it through your configured speech provider. Chat read-aloud also remembers playback speed per conversation in the current browser. Voice rendering handles escaped percent signs and degree notation more naturally, and Video Lesson's Edge fallback selects a voice matching the lesson language.

Easier provider setup and broader compatibility

Settings discovers available models after credential edits and lets you import several at once. New custom providers default to automatic protocol selection. NVIDIA NIM adds text retrieval embeddings with distinct query and passage roles; newer DashScope image models use their matching generation protocol, with 4K restricted to wan2.7-image-pro. Responses API tool arguments, structured output, replayed history, and Codex custom headers receive compatibility fixes.

Guided learning that keeps the context

Choose a model before a Mastery Path outline starts. Follow-up clarifications and missing parts are assessed with the explanation they complete, while stale card replies and duplicate retries are handled more carefully. Source quotation mismatches receive a bounded retry, and mastery navigation, board scrolling, and tutoring-memory recovery are more reliable.

More accessible conversations and retrieval

Korean is available as a partial interface translation, with English fallback for untranslated text. More labels and dates follow the selected interface language. Little Tutor threads appear in the command palette, document readers load on demand, and socket failures and cancellation recover more predictably. BM25 preserves dotted identifiers, document extraction stays off the API event loop, and notebook and favorite tools support read-only record retrieval.

Safer updates and bounded reasoning recovery

Package updates keep installed feature modules on the target core version and record launch failures instead of remaining stuck at restart. Detached launches avoid stale source-tree imports, and failed CLI-app rollbacks preserve backups. The shared agent loop can stop or make one explicitly bounded higher-budget retry when reasoning exhausts the output budget without a visible answer.

Upgrade Notes

• Full app: stop DeepTutor and back up the runtime home, then run python -m pip install --upgrade "deeptutor[full]==1.7.2". Keep the same DEEPTUTOR_HOME; verify with python -m pip check and deeptutor module list before restarting. Any additional feature wheels must also match version 1.7.2.
• Legacy installations: upgrades from 1.6.15 or earlier must uninstall the legacy deeptutor-cli package before installing the modular package family. Follow the migration commands in v1.7.0.
• Source installs: install requirements/full.txt, then python -m pip install -e . --no-deps. Stop active servers and CLI turns before changing modules.
• Docker: pull ghcr.io/hkuds/deeptutor:1.7.2 or :latest and recreate the container with the existing data volume.
• Desktop: use the matching installer or portable ZIP on this GitHub Release. Windows builds remain unsigned. Automatic desktop update discovery is not connected; the shell supports manually supplied signed core updates.
• NVIDIA NIM: select its embedding binding explicitly and re-index an existing knowledge base when changing embedding models. This binding supports text only.
• Speech and reasoning recovery: selected-text speech requires a working configured TTS service. Increasing the recovery output budget is opt-in and can increase latency and usage; the default retry budget remains unchanged.

Full Changelog: v1.7.1...v1.7.2