Artificial Intelligence (AI) is an area where there is a huge amount of innovation happening right now. So far, exploration of how to effectively leverage AI with DHIS2 is still in its early stages, but there is clearly huge potential to benefit DHIS2 users, such as streamlining data quality, facilitating data exploration, and powering predictive modeling, as well as some risks and ethical considerations to take into account.
To respond to this growing interest – and showcase some of the innovations that have come out of the HISP network and DHIS2 community over the past few years – we’ve decided to add a page to our website to share ideas and resources for using AI with DHIS2: DHIS2 & AI - DHIS2
This page highlights several projects that have been shared on the CoP and presented at the DHIS2 Annual Conference over the past few years, including:
If you’re interested in using AI with DHIS2, we encourage you to check out the new DHIS2 & AI webpage and share your questions and comments – and your own AI innovations – here on the CoP!
I have a dhis2 LLM App (Analytics Agent) im trying to build but im quite stuck and can’t seem to proceed.
The idea is officers in higher echelon positions of health and other organizations using dhis2, should simply login with their dhis2 instance, username and password and proceed to use the app like it were any other LLM, querying for health or organization data or responses regarding data stored in said instance, using natural language.
So the app is supposed to discern what type of query and produce the requisite response, the major queries being classified as “analytical“, where the response comes with a summary, key insights, the data itself, a variety of visuals changing as per change in parameter and finally foresights and recommendations.
For instance in this example, user queries “Show me malaria cases by district for the last 6 months“
However, i am facing challenges with follow up questions, the agent doesn’t seem to get the context from the response properly, among others.
Kindly review this.
I’ll be happy to share the code files / repo, with anyone interested in taking a deeper dive.
You may be interested in the AI Insights for DHIS2 app that Max linked to in the post above – it seems to do a similar analysis job, and may be useful to you!
Lastly, based on my (limited!) knowledge of using LLMs, to get the full conversation context with each message, you may need to send the whole conversation history along with each request so the model can process it all each time… maybe that’s what’s needed in your case
Good luck with the problem; hope to see you on Thursday!
you can find more details on the meetup page: 