search

coleam00 / bolt.new-any-llm

Prompt, run, edit, and deploy full-stack web applications using any LLM you want!
https://bolt.new
MIT License
3.88k stars 1.59k forks source link

Huggingface Models Integrated #314

Closed ahsan3219 closed 3 days ago

ahsan3219 commented 4 days ago

I have integrated the following Hugging Face models into our project for enhanced code and chat instruction capabilities:

  1. Qwen/Qwen2.5-Coder-32B-Instruct labeled as Qwen2.5-Coder-32B-Instruct (HuggingFace) provided by Hugging Face.
  2. 01-ai/Yi-1.5-34B-Chat labeled as Yi-1.5-34B-Chat (HuggingFace) provided by Hugging Face.
  3. codellama/CodeLlama-34b-Instruct-hf labeled as CodeLlama-34b-Instruct (HuggingFace) provided by Hugging Face.
  4. NousResearch/Hermes-3-Llama-3.1-8B for expanded capabilities.

This integration aims to support diverse coding and chat-oriented tasks within our project ecosystem. Let me know if you have any questions or need further details.

image image image

wonderwhy-er commented 4 days ago

Thanks for contribution. Remove yarn.lock as this project does not use yarn.

I will test it bit later.

ahsan3219 commented 4 days ago

Thanks

wonderwhy-er commented 4 days ago

Thanks

Will you remove yarn.lock? I can then test and merge if it works well.

I also looked at integrating dynamic list of models Filtered by code and its sorted by downloads to show popular models https://huggingface.co/api/models?filter=code&sort=downloads

But list of hugging face models is huge. Makes me think that we should only show top models and allow searching dynamically...

It just that with hardcoded models next week new one will come out and we will need to have new PR for it...

But yeah, we need some dynamic search option.

champ2050 commented 3 days ago

@ahsan3219, Kindly remove the yarn.lock so @wonderwhy-er can review and implement this. @wonderwhy-er, Agreed, showing top models and dynamic search would be more effective, especially with the release of new models. I'm excited about the integration of Hugging Face models, there are so many excellent options available, in my opinion.

But list of hugging face models is huge. Makes me think that we should only show top models and allow searching dynamically...

ahsan3219 commented 3 days ago

@ahsan3219, Kindly remove the yarn.lock so @wonderwhy-er can review and implement this. @wonderwhy-er, Agreed, showing top models and dynamic search would be more effective, especially with the release of new models. I'm excited about the integration of Hugging Face models, there are so many excellent options available, in my opinion.

But list of hugging face models is huge. Makes me think that we should only show top models and allow searching dynamically...

I have deleted yarn.lock file you can test it with npm

wonderwhy-er commented 3 days ago

Not sure why you closed it. Reopened, tested, works Merging

wonderwhy-er commented 3 days ago

Do you want to add any information or how to feature you in case some of youtubers will be mentioning you in ottodev update videos?

ahsan3219 commented 3 days ago

I will add other AI services and more AI models in this project so I have a lot to do to contribute in this open source Devine alternative....