# Analysis of a Gramps XML by an AI

**URL:** <https://gramps.discourse.group/t/analysis-of-a-gramps-xml-by-an-ai/6239>\
**Category:** Help\
**Tags:** hacks, websolutions, ai\
**Created:** [October 15, 2024, 2:41pm UTC](https://gramps.discourse.group/t/analysis-of-a-gramps-xml-by-an-ai/6239 "2024-10-15T14:41:44Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![PLegoux](https://yyz2.discourse-cdn.com/free1/user_avatar/gramps.discourse.group/plegoux/32/413_2.png) [@PLegoux](https://gramps.discourse.group/u/PLegoux)\
**Post date:** [October 15, 2024, 2:41pm UTC](https://gramps.discourse.group/t/analysis-of-a-gramps-xml-by-an-ai/6239/1 "2024-10-15T14:41:45Z")

</div>

Help or Idea?

Has anyone ever played with an AI to extract and format information by providing it with Gramps XML (or gedcom) and the right prompts? What is your experience?

I think I could format reports with this principle and with the right prompts. The idea is described in [a draft project](https://plegoux.notion.site/G-n-rer-mes-pages-Notion-en-utilisant-une-IA-qui-se-base-sur-un-export-XML-de-Gramps-120b33dc2f5e805a9576e14b79eaa5cc) (Google fr→en translation bellow) that aims to regenerate the pages of my Notion website (biographies) with an AI, or even to increase the content with context if possible.

* * *

## Automatically extract and format the content of my anscestors biographies on Notion

### ​​ **Goal and principles**

_What we are trying to achieve and the principles/constraints that must be taken into account_

### Background

- Extract an XML from Gramps (if I can avoid the Gedcom format, it’s better) on what I want to import into Notion
- Have an AI capable of analyzing it
- Have prompts to create a biography of each person from the information contained in the XML file.
- Same for families
- Same for surname lines in order to have a summary related to this surname
- Concerning the prompts, in addition to generating the previous information, they must be able to provide:
  - Factually by only taking the data from the XML file:
    - Have a Sources section to support the statements made.
    - Have a gallery when I have the necessary photos

  - Rather by looking at the side of sourced generative AIs (Liner, …):
    - Add relevant information to that coming from my data: for example pages on a particular profession or municipality
    - Format the texts around the person’s life according to their events
    - Have historical context on the places and times when people lived

- Determine what an AI can do with this data such as lifelines, create hotspots, and other interesting genealogical extracts

### Form

- Use one or more AIs
- Use [n8n](https://n8n.io/) for XML → AI(s) → Notion interactions
- The AI ​​must have an API allowing to submit the prompt, the data (all the XML? Bits?) and return the result in JSON format, if possible for the text in a format resembling Markdown
- Maybe subscribe to [BlueGPT](https://bluegpt.app/) which allows to interact with many different AIs from a single site and a single subscription

### **Final Vision**

_Expressed in the present tense, it describes the situation as if the project was fully completed, without any problems having arisen, and according to the greatest hopes that had been placed in it_

I extract my Gramps data from an XML and the pages [Généalogie de Patrice Legoux](https://www.notion.so/G-n-alogie-de-Patrice-Legoux-fb94f840291b4f06b4092295b22d55ba?pvs=21) are regenerated by an AI analyzing them. The AI ​​(or AIs) generate the blocks of the Notion pages that n8n assembles to publish the new pages. The AIs also generate content such as historical or geographical analysis related to the times and places of life of the members of the file.

---

<div class="post-metadata">

**Author:** ![GeorgeWilmes](https://avatars.discourse-cdn.com/v4/letter/g/c57346/32.png) [@GeorgeWilmes](https://gramps.discourse.group/u/GeorgeWilmes)\
**Post date:** [October 15, 2024, 5:17pm UTC](https://gramps.discourse.group/t/analysis-of-a-gramps-xml-by-an-ai/6239/2 "2024-10-15T17:17:04Z")

</div>

Interesting idea. I haven’t read the privacy policies of the various AI tools, but I wonder whether they would remember (on their servers) all of the data that you feed them, such that they might reuse it when responding to prompts from other users.

---

<div class="post-metadata">

**Author:** ![system](https://global.discourse-cdn.com/free1/uploads/gramps/original/1X/2ac1e712b7adf612e31ca08a55419f4cf37c0158.png) [@system](https://gramps.discourse.group/u/system)\
**Post date:** [November 14, 2024, 5:17pm UTC](https://gramps.discourse.group/t/analysis-of-a-gramps-xml-by-an-ai/6239/3 "2024-11-14T17:17:28Z")

</div>

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.
