Daniel

architecture: Jan Mobile

1

## Goal - Jan has a mobile client that runs local models

## Goal - Fix Local Engine issues - Focus on llama.cpp first ### Out-of-scope - Other engines: TensorRT-LLM, ONNX ## Tasklist Our issues boil down to 5-6 main cortex issues...

feat: Prompt user to download model, if no models are available

5

# Problem When there are no models available but user still has at least 1 old thread: - e.g. Factory reset (e.g. all models are deleted) - e.g. User deletes...

## Goal - Sub-task of #3690 - Jan should support new Model folder and model.yaml architecture from [cortex.cpp](https://github.com/janhq/cortex.cpp/issues/1154) - Jan should be able to migrate up/down, and deal with migration...

feat: Jan's Default Assistant can be customized

2

## Goal - Jan has a default assistant (which is initially called Jan) - User can customize Jan's default assistant (similar to how ChatGPT can be customized) - Name -...

feat: User can use Stop Words presets and add custom

3

# Problem - There are many users looking to enter multiple stop words - The "stop" section for configuring the model does not make it clear whether it should be...

## Goal - Jan supports TensorRT-LLM models ## Prereqs - https://github.com/janhq/cortex.cpp/issues/1152 - #3461 (tag-based TensorRT-LLM models) ## Tasklist - TensorRT-LLM Settings page - Hub has models with TensorRT-LLM tags (compiled,...

## Goal - We are able to have defect-free releases - We are able to double the size of team, without new hires breaking older code ## Key Questions -...

## Spec https://www.notion.so/jan-ai/Crash-Reporting-Telemetry-Stack-for-Cortex-Jan-9a920cc091bb447aa9db08b3477010e4?pvs=4 ## Tasklist - [ ] https://github.com/janhq/jan/issues/3119 - [ ] Pop-up Analytics opt-in after X mins of usage

# Overview - Should file an anonymous Bug on behalf of user, and give him/her the link - As this is public, we should 100% make sure that the logs...