●Change8

v5.11.0

Breaking Changes
📦 invokeaiView on GitHub →
⚠ 2 breaking✨ 16 features🐛 5 fixes🔧 2 symbols

Summary

This release introduces a major new model drop-down component and critical installation updates requiring the latest launcher (v1.5.0). Users must take specific steps to ensure a smooth upgrade, including backing up Form Builder state.

⚠️ Breaking Changes

  • If you are on v5.9.1 or older, you must update to the latest installer/launcher (v1.5.0) or the update may fail.
  • When starting Invoke for the first time after updating to v5.10.0, your Form Builder state will be reset, causing loss of unsaved changes. Users must save their current workflow before updating and manually reload it afterward.

Migration Steps

  1. If using v5.9.1 or older, download and use the latest installer/launcher (v1.5.0) to ensure a successful update.
  2. If the installation fails, retry the installation with **repair mode** enabled (tick the checkbox on the **Review** step of the install) to resolve Python environment conflicts.
  3. Before updating, save any unsaved workflow changes, as the Form Builder will reset upon first launch after updating.

✨ New Features

  • New and improved model drop-down component implemented (currently only for the main model drop-down).
  • Added button to reset an existing Hugging Face (HF) token in the Model Manager tab.
  • Support for FLUX LoRAs trained in `invoke-training`.
  • Nodes that output images (including image collections) should now always update the gallery.
  • Internal: Optional `output_meta` field added to `BaseInvocationOutput` for future metadata attachment.
  • Internal: Support code for generation via Imagen3/ChatGPT 4o (currently unavailable in Community Edition).
  • Support for Non-Standard LoRA Format.
  • Display credit column in queue list if `shouldShowCredits` is true.
  • Add credits to queue item status changed events.
  • Support for `usage_info` and ChatGPT/Imagen3 model fields.
  • Feature to disable generation with API models changed to specifically target `chatGPT4oModels` only.
  • Use the max for iterations passed in.
  • Refetch queue list on mount.
  • Easier way to override Whats New display.
  • Support for custom error toast notifications.
  • Support for bottleneck flag for nodes.

🐛 Bug Fixes

  • Fixed issue where drag-and-drop did not scroll when used inside a scrollable container (e.g., many layers or form builder elements).
  • Fixed a toast typo.
  • Fixed an unhandled exception while registering the hotkey handler by ensuring `selectModelsTab` hotkey data is always added.
  • Fixed a deprecation warning in `get_iso_timestamp`.
  • Added prompt validation requiring at least one character for Imagen3 graphs.

🔧 Affected Symbols

BaseInvocationOutputget_iso_timestamp