Frederik Wouters
The seed was planted earlier but today it was communicated for the first time. People at Calibrate can ask all their questions to a bot that has waded through the entire (internal) documentation.
Meet Booksbot, our intern searchbot
Here you can see an example interaction with Booksbot.
You just ask a question and the bot answers with the information from your internal documentation. As an attachment, the bot also displays the references in the form of citations (where the documentation comes from), always handy to click through immediately.
'Tis really pleasant that people can now find internal information even more easily. Only the challenging questions still reach people.
Resources
We started with two internal sources of information.
- Slack
- Our Bookstack (all our internal information).
But in time we may want to add resources to this.
- Sharepoint
- Zendesk
- Confluence
- Gitlab
- Google Drive
Parts
Booksbot consists of several components. It is an Open source AI orchestration. It uses Python, a Vector Database and 2 AI models (an embedding model and an LLM or language model). We can customize the prompts ourselves so that gives a lot of possibilities. The official term is RAG or Retrieval-augmented Generation, also called generative AI colloquially.