jupyter-ai
jupyter-ai copied to clipboard
jupyterlab
A generative AI extension for JupyterLab
[Groq](https://groq.com/) has tremendous inference speeds (280 tokens per second for Llama 3 70B and 877 tokens per second for Llama 8B0. It would be amazing to get support for this...
## Description The `help` message suggests https://github.com/jupyterlab/jupyter-ai/blob/e27293cf86fa37b2ea20de6b481ad989e209933a/packages/jupyter-ai-magics/jupyter_ai_magics/providers.py#L503-L505 but passing it in this format results in: ``` pydantic.v1.error_wrappers.ValidationError: 1 validation error for HfHubProvider __root__ Repo id must use alphanumeric chars or...
### Problem Hi- I’m working with [Clay](https://clay-foundation.github.io/model/index.html), a foundation model for Earth Observation data. I’m exploring what an integration with Jupyter-ai would look like, but have a couple questions due...
### Problem I have set the parameters for Azure in the configuration page. Language model: Azure OpenAI Deployment name: gpt-35-turbo-16k Base API URL: https://xxxxxxx.openai.azure.com/ API version: 2023-06-01-preview API Keys: xxxxxxxxxxxxxxxxxxxxxxx...
## Description As noticed by @krassowski in https://github.com/jupyterlab-contrib/jupyter-chat/issues/15, the releaser bumps the version 2 times in the workflow, with the current settings. Related issue: https://github.com/jupyter-server/jupyter_releaser/issues/567 Some logs from a recent...
### Problem When websocket gets disconnected, the log console gets spammed with messages every second:  Making it hard to debug other issues ### Proposed Solution - Use some function...
## Description When websocket gets disconnected we see: ``` Uncaught TypeError: Cannot read properties of undefined (reading '_reject') at reject (jlab_core.0c882064cb12a87cda90.js?v=0c882064cb12a87cda90:1:1731169) at we._onClose (392.e7fabd325dafbe2024f7.js?v=e7fabd325dafbe2024f7:1:37258) at i.onclose (392.e7fabd325dafbe2024f7.js?v=e7fabd325dafbe2024f7:1:37860) ``` ## Reproduce...
## Description When running the `/learn` command, users can provide an absolute path relative to the host operating system's root directory. For example, on macOS, the path `/Users/jsmith/foo/bar` is valid,...
### Problem `DefaultInlineCompletionHandler` makes some choices which may be sub-optimal when using a models tuned for code infill tasks. For example, the reliance on langchain template system here: https://github.com/jupyterlab/jupyter-ai/blob/4722dc73ae226f21108812a16fe2a5496aac9343/packages/jupyter-ai/jupyter_ai/completions/handlers/default.py#L45-L47 makes...
## Description Chat memory is reset when selecting new LLM provider. https://github.com/jupyterlab/jupyter-ai/blob/3bfce328e3b6f730d05faa68ca9b6d6434b5fdec/packages/jupyter-ai/jupyter_ai/chat_handlers/default.py#L34-L36 This is due to `self.memory` being reassigned in `DefaultChatHandler.create_llm_chain` which is triggered when LLM provider is changed. ##...
