How to implement Q&A against your documentation with GPT3
Jan 16, 2023
- Run a text search (or a semantic search, described later) against your documentation to find content that looks like it could be relevant to the user’s question
- Grab extracts of that content and glue them all together into a blob of text
- Construct a prompt consisting of that text followed by “Given the above content, answer the following question: ” and the user’s question
- Send the whole thing through the GPT-3 API and see what comes back
I’ve been calling this the semantic search answers pattern.