I hope this message finds you well. I’m reaching out to propose the integration of GPT-3.5 into our forum’s AI bot. With OpenAI making GPT-3.5 accessible without registration, we have a unique opportunity to enhance our forum’s user interaction and content generation capabilities.
By incorporating GPT-3.5, we can significantly improve our AI bot’s response accuracy and user engagement, fostering a more dynamic and informative forum environment.
I believe this integration will not only elevate our user experience but also showcase the potential of AI in community engagement. I’m excited to hear your thoughts and am grateful for any contributions towards making this integration a reality.
Thank you for your consideration and continued support.
Firstly, I’d like to express my appreciation for the Discourse Chatbot plugin you’ve developed. It’s truly an innovative tool that enhances the interactive experience within Discourse forums.
I understand that the chatbot’s current default setting is GPT-3.5, which operates based on API interactions. While this setup offers substantial benefits, I’m interested in exploring the possibility of integrating GPT-3.5 into the chatbot without relying on the API. The aim is to streamline the integration process and potentially enhance the bot’s responsiveness and accessibility.
Would it be feasible to implement a version of the chatbot that utilizes GPT-3.5 directly without the API dependency? Such a feature could offer an alternative for environments where API interactions are constrained or where a more autonomous setup is desired.
Thank you for considering this suggestion. I’m keen to hear your thoughts on the feasibility and potential impact of such an integration. Your efforts in developing and refining this plugin are greatly appreciated, and I’m looking forward to its continued evolution.
thanks for your kind words, it was a lot of work (and shout out to those that have supported the project!).
GPT 3.5 , though no longer the cream of the crop, is a very large model and requires a significant amount of processing power to run - so much so you would likely need resources of a significant part of a datacenter. I do not believe it is feasible to run such a large model on your own infrastructure. You would be overwhelmed by the cost. GPT 3.5 is also not “open sourced” and therefore it is not an option.
That said, you can explore options with open source models via the chatbot’s proxy settings.
Open source models can be browsed on the Ollama directory:
I’ve been looking at a few proxy projects, some that you can run locally and some in the cloud like Vercel.
e.g.
You can also explore other proprietary models like Gemini via proxy:
Thank you very much for your detailed response and for shedding light on the practical aspects of integrating GPT-3.5 into our infrastructure. Your explanation provides a clear understanding of the challenges associated with running such an advanced and resource-intensive model.
I greatly appreciate the suggestion to explore open-source models through the chatbot’s proxy settings and will definitely look into the Ollama directory for potential options that could fit our needs.
While we understand the limitations with GPT-3.5, we remain hopeful that in the future, there might be more accessible ways to integrate such powerful technologies directly, perhaps through an embedded proxy that could simplify the process and make it more cost-effective.
Thank you once again for your support and the valuable work you’re doing with the Discourse Chatbot plugin. Looking forward to any future developments in this area.