Customize ChatGPT for creating Bot for organizations internal wiki

ChatGPT
1

HI All,

We are planning to create chatbot which serves the purpose of wiki /confluence we internally use.

We need to feed the our wiki data to ChatGPT and expose to internal users using bot application which can call ChatGPT api to get the data based on user input.

Below are my queries .

1, How do we feed custom data to chatgpt
2, if we feed our internal data will it become publicly accessed ,if how do we make it private in chatgpt
3, is there any helloworld examples for feeding data to chatgpt? please share the reference ? is it available in c#?

have a great day ahead

2

Hope this helps.

1 Like
3

Hi @ezycoder

@PaulBellow is correct, chatGPT isn’t available on the API as of now.

However if you are using chatGPT to refer to GPT-3 models, then it’s doable with embeddings

4

Hi,

You will need to break your knowledge base into many small chunks and use embedding endpoint to calculate an embedding vector for each chunk. You will then need to use a vector database like pinecone to store both the vector and the original chunk of text related to the vector. To use this in a chatbot you will need to develop a chatbot prompts that answers questions based on context. As the user talks with the chatbot you will need to embed their message and search against the vector DB. It will return N similar text chunks. You can then add these text chunks into the prompt as context and then when you run the prompt the response will answer based on the context.