[Question]: building custom GPTs using embedding system

[chminsc]

Describe your problem

After create a knowledge base in Ragflow, is there a way to use openai's GPTs to query the knowledge base? any api that I can use? on that way I can leverage chatgpt's capabilities without extra paying

[KevinHuSh]

Set a valid API key in Model provider.

[chminsc]

Set a valid API key in Model provider.

I'm saying, using GPTs function of chatgpt's online website, to query ragflow's knowledge base.

[KevinHuSh]

Do you have any document or reference there about using GPT functions?

[chminsc]

Do you have any document or reference there about using GPT functions?

here are some apps: https://chatgpt.com/gpts here is the official introduction of action/schema used by GPTs: https://platform.openai.com/docs/actions/introduction

here is a very good example: https://github.com/openai/openai-cookbook/blob/f237586d8001acc1fa1526de4aa86836ed3fbf40/examples/chatgpt/rag-quickstart/pinecone-retool/gpt-action-pinecone-retool-rag.ipynb

so the basic idea is to use GPTs to query ragflow for a question, and return the search outputs(chunks?) to gpt, and gpt will give the final answer. same logic as ragflow, but use remote llm(not api). it's doing the reverse way - the remote server will query local ragflow service as a vector server.

[chminsc]

so I think expose a vector query api is enough for this to be done? That I can create a action on GPTs for query

[KevinHuSh]

Thanks for sharing.

I think RAGFlow has those APIs. Please refer to this.