Add in-browser LLMs and set Qwen2.5 0.5B as default. Add Safari support.#411
Merged
Conversation
…overcome Safari resize bug.
* Fix #406: Update deprecated HuggingFace API endpoint Replace deprecated api-inference.huggingface.co/models/ endpoint with api-inference.huggingface.co/inference-endpoint/ to resolve 410 Gone errors. * Fix #396: Close context menu before executing menu item actions The right-click Duplicate Node action failed on some systems (e.g., Windows 11 LTSC) due to a race condition where Mantine's useClickOutside handler would close the menu on mousedown before the item click handler could execute. The fix closes the context menu explicitly in the onClick handler BEFORE calling the action handler. This ensures proper event ordering on all systems. Also applied the same fix to 'Favorite Node' and 'Delete Node' menu items for consistency. --------- Co-authored-by: Ernest <ernest@example.com>
Add MiniMax M2.7 and M2.7-highspeed models as a native LLM provider, following the DeepSeek/OpenAI-compatible pattern. Includes: - NativeLLM enum entries for MiniMax-M2.7 and MiniMax-M2.7-highspeed - LLMProvider.MiniMax with provider detection and rate limiting - call_minimax() via OpenAI-compat API (https://api.minimax.io/v1) - Temperature clamping (min 0.01) per MiniMax API requirement - Settings schema with model selector, temperature, system_msg, top_p, max_tokens, stop, presence/frequency_penalty - UI menu group with both models - Flask env var mapping for MINIMAX_API_KEY - Fix: call_chatgpt now respects custom API_KEY param (skips OPENAI_API_KEY check when a custom key is provided) - 25 tests (21 unit + 3 integration + 1 conditional) Co-authored-by: PR Bot <pr-bot@minimaxi.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This adds in-browser LLMs using WebLLM. LLMs will one-time download to your browser cache, and then run entirely locally. We included two options by default for their size and on-device performance.
The in-browser quantized Qwen2.5 0.5B is now the default model when PromptNodes are added. We think most devices today will be able to run this model, which is ~1GB download and once downloaded, runs fast.
We also unblocked Safari support—it is now 3 years since ChainForge originally released, and the current version of Safari has gained support for the features blocking its access previously. We needed to tweak some things regarding CSS and resize handlers especially (added a new
ResizeHandlerclass for cross-browser consistency). If you still have problems, please let us know.