Logging conversations with GPT in database

Hello - new to the openai community. Out of curiosity, does OpenAI have plans to give ChatGPT the capabilities to store conversations in a database? I am building a personal GPT and I’d like to improve based on previous interactions.

Is something like this in the works? Are there reasons not to implement such functionality. Any insights are greatly appreciated.

Best regards,

Peter

Spying on GPT conversations users thought were private is not currently supported unless you are building an action API.

2 Likes

Hello and welcome!

You can do an export of your chat history and view your chat history.

If you use other devices you’ll need to do an export on those as well!

To do this click on settings

Then click on data control and press export data

Happy chatting :smiling_face:

@ptumulty923 what kind of information are you looking to acquire from these logs? What are you hoping to improve? I’m working on something that might be helpful to you

1 Like

I’ve currently got about 750 conversations archived in a tiddlywiki; it’s a bit slow to load and cumbersome to process, and involves constantly exporting chats from open AI. I haven’t quite been ready to archive all my chats, which I think would be a way to identify them as “exported” and speed things up…

Structurally, each prompt and each response has its own tiddler, which will allow future enhancements such as searching tagging or annotating specific prompts or responses rather than full conversations.

The process involves using open AI to generate a json file containing the archive, running a python script on the json file generated by open AI to generate a json file that can be imported cleanly into tiddlywiki, importing the json file into the tiddlywiki, and using a template file to render conversations and messages as desired

If you are using the API and would like to trace the code. You can have a separate method which has all the inputs and outputs you’d like to trace. I personally use Langfuse for tracing. In that way you can pick the params you’d like to observe…