While the interface of the Kairntech Chatbot is intentionally kept simple in order to allow for intuitive use, there are rich configuration possibilities in the backend.
The Kairntech Chatbot connects to the Kairntech Server via API. Its behavior can be customized via Kairntech Studio user interface.
How to build a Kairntech Chatbot?
- Create a Question-answering project in Kairntech Studio – see here
- Import documents (via API or manually – see here)
- The uploaded documents are automatically converted, segmented, vectorized and indexed with a default configuration. A default LLM is also pre configured (GPT-4o-mini).
- An alternative could be to use our Sharepoint connector to automatically ingest content into the project
- Connect to Kairntech Chatbot with your credentials
- Your chatbot is ready!
How to customize a Kairntech Chatbot?
- Go to the Processing menu
- Select the LLM that will act as the agent using the yellow star.
- Note that only a limited number of LLMs are compliant with the chatbot implementation (GPT family including gpt-OSS, DeepSeek V3, Llama 3.3, Gemini Flash 1.5…)
- Go to the Messages menu
- You have access to the agent Prompts.
- you can edit it and add a new prompt text to supplement the default one
- Let’s study the default prompt from the “System Prompt” in details:
The prompt gives the LLM clear instructions about how the Chat should behave. As in many other places, the system offers reasonable defaults that have been found to work with a variety of LLMs that are accessible via Kairntech.
- You have access to a number of “AI agent tools” and can edit each of them through simple text messages:
- “Search in documents” is the predefined tool that attempts to answer a user’s question by accessing relevant documents from the stored content (Retriever function from a RAG scenario for instance).
- “Classify question” determines the user’s intention as to whether the LLM should use its general knowledge or activate a specific tool (such as document search).
- “Define chat parameters” allows the user to switch in debug mode. For instance the agent picks up user requests such as “Switch on the debugging mode” and changes the mode of the Chatbot to allow inspection of the internals of a chat.
