ShareGPT

• 

  Collaboration in ChatGPT with shared context, like play tic tac toe with a friend (each in their own chat)

• 

  Save conversations and formulas with ShareGPT

• 

  Run formulas, your own or from other users with ShareGPT

Inspiration

• ChatGPT is not a collaborative environment - yet!
• New ChatGPT session do not have any context, you need to build from scratch each time
• Copy pasting all those prompts from Twitter threads is tedious

What it does

• Adds universal save and load functionality, shared memory and context even between multiple instances (shared clipboard or database)
• Turn those recordings into more powerful formulas (think functions with parameters) and share them (how about a Github for ChatGPT?)
• Mix it with your own plugin, e.g. spawn agents with a shared memory from the main conversation, store your project information

How we built it

• ChatGPT plugin with FastAPI server and database (MongoDB, experimental Weaviate) deployed on Vercel

Challenges we ran into

• Limitations of ChatGPT API, limited access to chat history, single prompt for response

Accomplishments that we're proud of

• Simple idea, massive potential - ever played tic tac toe with your friend in ChatGPT?
• Amazing prompt engineering to ensure a great user experience and store summaries, build functions and without ever leaving ChatGPT
• Enabling so many use cases and adding a universal building block to take ChatGPT and other plugins to the next level

What's next for ShareGPT

• Version control (manage the history, go back in time, collaborate with others, merge)
• Hierarchical memory (load details on the fly)
• Link sharing with social previews (start the social movement!)
• Asset management (text is only the start, think images, files, vectors, ...)
• Chained workflows (formulas with multiple steps, or even multiple formulas)

Built With

• chatgpt
• fastapi
• openapi
• python
• vercel