jan
jan copied to clipboard
janhq
Jan is an open source alternative to ChatGPT that runs 100% offline on your computer. Multiple engine support (llama.cpp, TensorRT-LLM)
## Objective - WIP spec ## Resources - Apple Ferret - Need to support MLX - Support Quantization - https://www.reddit.com/r/LocalLLaMA/comments/18oke4y/apples_mlx_framework_adds_quantization_support/
## Resources - https://www.reddit.com/r/LocalLLaMA/comments/18i5imw/windows_ai_studio_offical_local_finetuning_gui_in/
## Description - Inference & system information extensions are depends on Node runtime, we should move these deps to /core so extension would not run any piece of code on...
**Problem** - when we ship breaking changes, sometimes the `/models` `assistant` folders need to be overwritten/deleted/purged - I have to manually reset app right now - Edge case: #1072 -...
# Problem Windows users encounter elevated CPU usage (ex: #887), prompting the need for hardware optimizations in Nitro/Jan across various Windows architectures (CPU, GPU). ### Success Criteria - Achieve compatibility...
**Motivation** - People dont always know how to prompt - People want a prompt db - People want to save /configure their own custom prompts - Example: People want a...
**Describe the bug** A clear and concise description of what the bug is.  stopping the execution fails (note CPU load remains 100%)  subsequent threads broken  opening a...
**Problem** Jan UI should have a reload model button for people like me who constantly break stuff -- opening a new thread with the same model doesn't reload it so...
**Describe the bug** when model is missing from model folder but json exists jan ui says model active while nitro doesn't load the model, leads to confusing fail to fetch....
**Pages** - Troubleshoot/Fail-to-fetch - Docs captures new features from Jan Nightly Build that pop-up the reason of error **Success Criteria** Help users easily detect the reason of fail to fetch...
