3,146
edits
Changes
From Gramps
→How does it all work?
== How does it all work? ==
The ChatWithTree is build built in such a way that it sends off a list of "tool names" to the LLM, and based on the message send, will determine what tools it needs to respond. The tools in the respond execute an interaction with your Gramps database, and the results are then send back to the LLM again, so that the LLM can digest the information for an answer.
This interaction between tool calling, thinking of the model, is shown in the ```yellow balloons```.
