Ask HN: Internally-Hosted ChatGPT Proxy?

1 point

3 years ago

I'm investigating options for hosting a corporate-internal ChatGPT UI that calls the OpenAI APIs. A reason to do so is that OpenAI has stated it will not use data submitted by the API to train or improve future models[1], while it will use prompts, responses, etc to improve services in their "consumer" products.

Ideally this would support a lightweight proxy for the API, some kind of login, and some ChatGPT-like features like chat history (serving the right history to the right user :-)

Has anyone else pursued this? Right now it looks like there are a handful of alternative UIs, like chatbot-ui[3] and chatgpt-ui[4]. I haven't run into an API proxy supporting some of the detailed features like streaming, though it wouldn't be too hard to build.

[1] https://openai.com/policies/api-data-usage-policies

[2] https://help.openai.com/en/articles/7039943-data-usage-for-consumer-services-faq

[3] https://github.com/mckaywrigley/chatbot-ui

[4] https://github.com/dvcrn/chatgpt-ui

2 comments