EchoGrove v1.0.4 Beta Changelog

Release date: 01-06-2026

This build is mainly a stability and setup release. It focuses on AI provider reliability, first-run setup, chat feedback while models are thinking, and a few appearance/profile cleanups.

What Changed

AI provider fixes

Fixed AI Assist failures with OpenAI and Gemini keys where some requests could return 401 even when the provider was configured.

Fixed Gemini chat hanging on Thinking... by using a non-streaming Gemini response path. Gemini replies should now complete normally.

Chat now uses the model selected for the configured Chat tier, and Grove passes that selected model into the chat context so the assistant can identify the model it is actually using.

Provider model handling was consolidated so OpenRouter, OpenAI, and Gemini all go through the same tier/favourite model system.

AI chat feedback

The old global top-left Thinking... banner no longer appears for chat.

Chat now shows its thinking state inside the assistant message bubble.

While waiting, the message displays elapsed time, for example Thinking... 3s.

After about 20 seconds, Grove adds a note that reasoning models can take longer, so a slow model does not look frozen.

First-run author setup

Grove now asks for your author profile on first launch if no profile exists.

The setup asks for your name and starting highlight colour.

The colour chosen during setup now drives Grove's Accent/highlight colour.

The separate Profile colour picker was removed from Settings because Accent is now the single place to change that colour later.

Settings and appearance

Settings > Appearance now has a small Profile section for editing your Grove profile name.

Profile wording was cleaned up to remove old collaboration/comment attribution language.

Spacing in the Profile section was adjusted so it reads better at default UI zoom and larger UI scale.

Dark-mode selection/highlight visibility was improved.

Manual and help text

The generated in-app manual was updated for the new author setup and profile wording.

References to the old comment-attribution behaviour were removed from the first-run setup documentation.

What Beta Testers Should Try

Open the installed Grove app and confirm existing projects still open normally.

In Settings > AI, switch between OpenRouter, OpenAI, and Gemini if you have keys available.

Send a normal AI chat message using Gemini and confirm it completes rather than staying on Thinking....

Ask chat what model it is using and confirm it matches the selected Chat tier model.

Open Settings > Appearance and confirm Profile name editing saves cleanly.

Change Accent colour and confirm highlights/buttons follow the selected colour.

If testing a clean install/profile, confirm first-run setup asks for name and colour, and that the chosen colour becomes the Accent colour.

Verification Before Release

Production build passed.

Windows release installer build passed.

Vitest passed: 30 test files, 350 tests.

Installer

Windows installer produced:

EchoGrove_1.0.4_x64-setup.exe

Before installing this beta, uninstall the previous Grove version from Windows first.

Important: during uninstall, do not tick Delete the application data unless you deliberately want to remove your local Grove profile and current project/library data. Leave that box unticked to keep your existing profile and projects.

As usual, it will take a while to be updated, notarised and signed by Apple for the corresponding MacOS installer.

Known Notes

The thinking-level selector remains parked as a future enhancement.

DM me if you would like to test EchoGrove.