-
It actually works (in Firefox)
-
Adding WebLLM backend as an option.
-
Fully working version at last.
-
Disabling model discovery for now.
-
Slash-loading regardless of lookup.
-
Hide menu on escape or click the item.
-
Slash menu style tweaks.
-
Slight tune for the menu.
-
Simpler menu.
-
Slash: attempt to fix.
-
Faster fetching of model list (with caching).
-
Reflect model count.
-
Model load fixes.
-
Worker loading model list tied into the UI.
-
Followup on worker.
-
Core refactoring of list-chat-models, and a plan for the follow-up.
-
Implemented the worker methods.
-
Trying to improve filtering.
-
Async loading of the slash/models.
-
Cache model list for network failures.
-
Tweaking model list a little.
-
Caching slash-loaded models to use.
-
Handling early output better.
-
Improved model filtering.
-
Better icons for model loading.
-
Slash menu now lists the models.
-
Loading models into slash.
-
Enabling more natural styling and slash menu.
-
REmove dodgy styles.
-
Crepe input.
-
Preparing to slash commands (for list of models and switch model).
-
Surface version in the document.title.
-
Fix to autoscroll output.
-
Layout tweaks.
-
Report transformers.js env.
-
Better fallbacks.
-
Use auto for the backend.
-
Collect whole output though.
-
Report version when loaded.
-
Lighter model.
-
Lighter model.
-
Wire the prompt across from UI to worker to LLM.
-
Updating to handle prompt queries better.
-
Wiring some of the models to be used.
-
Wiring more support for loading models.
-
Introducing a worker to handle transformers.js without blocking.
-
Basic prompt reply.
-
Auto-focus on input.
-
Update to enter key handlung.
-
Enter key to type.
