I finally got around to exploring LLM + tools with gramps. The basic idea is to make available a series of functions that can access gramps object data (the dictionary format used in JSON API, etc). I’ve added 13 functions/tools so far, things like get_person(), get_person_death_place(), etc. This …

Having the whole personal database uploaded to the cloud sounds a bit risky for privacy reasons, that is, excluding living persons and all. The nice thing about running the chatbot locally is that locally the right tools are selected to query the database. Extending these tools seems the challenge. …

can LLM be trained on the Gramps wiki That’s a different question from interacting with a local gramps database. Training an LLM takes huge amounts of processing time, and the context windows of current LLMs were small until Gemini hit the 1 million token barrier. I usually just use the famous s…

[image] Hace: The nice thing about running the chatbot locally is that locally the right tools are selected to query the database. Extending these tools seems the challenge. As an alternative, I am trying to play with a Client Challenge Few recents mode…

[image] Hace: Have a look what I added for the get_child_in (families) - and the example chat. I am also surprised that some questions are too hard for the model, like “who is the brother of the home person” - would say with the amount of tools it would be able to work that out, but that is not …

The remote gemini flash-2.5, very strong model. The local machine is very old, but that is fine, it can execute each gramps query at the expected speed of this old machine. I the model usually does select the correct tools, and it also says what kind of tools might be missing. That is, A reply onc…

Making progress on a grampschat server: [image] TODO: Find an ollama model powerful enough to handle tools (above is using openai/gpt-4o-mini) Find a machine in the cloud that is powerful enough to run the local model Figure out cost of server Then lots of little things: how to upload databa…

Hi @dsblank finally managed to add a “find_people” tool, so that we can search on people by name in the database. I needed to also restrict the AI Model from looping, and added a final prompt after a few tries, not providing any tools anymore, to force an answer from the local running modal. You m…

Thanks for the update. FYI, the method can be pointed to using the github URL: genealogychatbot/chatbot.py at master · MelleKoning/genealogychatbot · GitHub I’m going to try to make such a query function much more efficient. But what you have should work ok for small trees.

True about small trees. When a tree contains tens or hundreds of thousands people this won’t perform. Also notice the summarization of the results with “desired fields” As I found that even a result on a lastname with about 25 matching people, resulted in a huge json addition to the context window…

That subroutine seems to ignore some of the common name options. Will it make the chatbot too formal to recognize all the possibilities? For instance, when searching for people who colaborate online, I might only recall their online name. So if I search a Gramps tree for kku, I’d want to find Kari …

LLM + Gramps = a useful chatbot

Development

Topic		Replies	Views
GrampsChat Addon for 6.0 Ideas third-party-addon , ai	21	649	March 24, 2025
Gramps Assistant Development ai	46	819	April 12, 2026
Chat with your genealogy? Gramps Web	4	650	January 1, 2026
ChatWithTree gramplet addon Development third-party-addon , ai	6	381	December 30, 2025
Talk to your tree! Gramps Web ai	46	965	January 25, 2026

LLM + Gramps = a useful chatbot

Related topics