DocsBot AI

Yes! Unlike traditional search, DocsBot inherits GPT's understanding of any language found on the Internet. So even if you train from your documentation in English, it can still find answers and respond in the language of the question!

It's a bit technical, but here is a brief overview. We use OpenAI's embedding and ChatGPT APIs, as well as vector databases to store our index. All ingested documentation is cleaned up and divided into smaller chunks and labeled by source. We then generate a vector representation of each chunk and store it in our vector DB index. When a user asks a question, we convert it to an embedding, and perform an advanced semantic & keyword search for closest matches to the user's query. Then we take the most relevant chunks, included them as context along with the original question, and use the ChatGPT API to generate a response in markdown format that we then convert to HTML and display to the user.

Access to the capable GPT-4o mini or GPT-4.1 nano models is included with your plan question limit, and works well for most use cases. To access other models such as the most powerful GPT-4o or GPT-4.1 you need to provide your own OpenAI API key. Instead of simply reselling API access at a premium, we think it is much fairer to have you only pay for the usage you need. We also think it is important for you to have full control and ownership of your data. This also prevents rate limits from causing outages for you if another of our customers has a spike in usage. Each question uses <$0.008 of credit, so costs are minimal. We store your API key securely with AES256 encryption, and it is only used to make calls to OpenAI on your behalf.