DHIS2 AI Assistant Chrome Extension (DHIS2 AI Metadata automation tool)

Development - Développement
,
1

DHIS2 AI Assistant Chrome Extension (DHIS2 AI Metadata automation tool ) — Now Available on the Chrome Web Store

Hi community

I’m excited to share the DHIS2 AI Assistant, a Chrome extension that brings an AI-powered side panel directly into your DHIS2 instance. It’s now available on the Chrome Web Store and works with any DHIS2 server you’re already logged into — no passwords stored, no separate login required.

:link: Install it here: https://chromewebstore.google.com/detail/dhis2-ai-assistant/fodjnbajmgkinpcebbekilimbkgoldoa

What it does

The assistant lives as a side panel inside Chrome and talks to DHIS2 through your active browser session. It automatically detects what page you’re on — a tracker program, a dataset, a visualization, a map — and narrows its capabilities to what’s relevant. You interact in plain English; it handles the API calls.

It comes with 22 specialized tools covering:

  • Analytics queries, event counts, enrollment checks, and line listing guidance.

  • Full metadata authoring: tracker programs, program stages, data elements, option sets, category combinations, datasets, sections, sharing, org-unit assignments, icons.

  • Program rules and program indicators — audit, create, bulk-fix expressions and conditions.

  • Visualizer and Maps context — load and explain any saved chart or map layer.

  • Auto-backup before every destructive operation, with one-command restore.

  • Export any response as HTML, Word, XML or JSON.

What makes it different from other AI tools is context-awareness: the extension travels with you as you navigate DHIS2 and reads your current page context automatically, so you don’t have to manually tell it what program or dataset you’re working on.

You can build entire programs with stages, data elements, attributes, option sets, program rules, OUs and more, just by tell it what you want in plain English.

Getting Started

1. Install the extension
Click the Chrome Web Store link above and install.

2. First-time access on a DHIS2 server
When you open the extension on any DHIS2 server for the first time, it will ask:

“Allow access to this DHIS2 server?”

Click Allow. A Chrome permission dialog will then appear — click Allow again to grant the extension access to your DHIS2 session. This is required for the assistant to make API calls on your behalf.

3. Configure your AI provider
Click the gear icon to open Settings and choose how you want to power the AI:

  • Ollama (local, no API key needed): Set the API Base URL to http://localhost:11434/v1. Important: if using Ollama, you must use a strong model with solid agentic/tool-calling capabilities — the assistant runs multi-step reasoning loops and calls multiple tools per turn. A weak or small model will produce poor or incomplete results. Good options: Kimi K2.6, GLM 5.1 or similar frontier-class local models.

  • Cloud provider: Choose any OpenAI-compatible provider, enter the model name, and paste your API key. Providers include:

  • OpenAI

  • Anthropic

  • Google Gemini

  • OpenRouter

  • Fireworks AI

  • If you want to use a proprietary LLM, I advise you to use (gemini-3.5-flash), as its affordable proprietary model, yet it’s fast and relatively strong with tool use.

  • If you want to use open-source models with cloud providers such as Fireworks AI, try Kimi K2.7 Code. This newly released model is strong at agentic tasks, very affordable and fast.

4. Navigate to DHIS2 and start asking
Log into your DHIS2 instance, open the side panel, and ask in plain English:

  • “How many enrollments does this program have this year?”

  • “Create a new tracker program for antenatal care with 3 stages, then give it the program details”

  • “Audit the program rules in this program and fix any broken conditions”

  • “Explain this visualization and show me the data behind it”

This was built by a HISP MENA developer for the DHIS2 community. Feedback, bug reports, and contributions are very welcome.

Here is the GitHub repo: GitHub - AbdelatifAbualya/dhis2-ai-extension: AI-powered side panel for DHIS2 (Chrome MV3). 22 tools — analytics, tracker writes, atomic metadata authoring (programs, datasets, category combos, sharing), rule/indicator audit + repair, backups. Any OpenAI-compatible provider; defaults to local Ollama, no API key. Authenticates via active DHIS2 tab session. · GitHub

4

This looks like a fantastic tool, the context-awareness as you navigate between programs and datasets is a smart touch. Does it work well with DHIS2 2.39/2.40 instances, or are there version limitations to be aware of?

5

@Abdelatif-Abualya
This is a cool innovation. I can’t wait to keep testing this.

6

Hello Raphael, it will work with DHIS2 v2.40+

7

As this is agent-based we can’t use Ollama, it doesn’t seem to make much sense. Running a larger model can significantly slow down the machine, while using a smaller, less capable model increases the chances of errors.

Therefore, using an API key seems like a more practical approach, as Ollama can be too slow for agent-based workloads.

Is this project open source? Also, which framework is it built with Spring AI or a Python-based framework?

If it is open source, could you please share the GitHub repository, @Abdelatif-Abualya?

8

Hello @Mohd_Nawaz, You can use Ollama, but as you mentioned, LLMs that are good at agentic tasks/ tool calling are large, to be able to use it with Ollama you need a powerful hardware. So for many, using it with LLMs through the API is their best option, and its for this reason, the LLM is blocked from seeing any data other than aggregate data, and metadata.

I will share below the GitHub repo URL, the extension was built with Javascript, and it completely open source with MIT license, you can use it, edit it or share it.

9

I really appreciate this part:

I think that the extension can be extended into doing many things in addition to the metadata automation, and I appreciate how you handled the AI usage with consideration for LLM (sustainability and privacy)…

I can’t find a Github repository, is it available for the public? never mind, I saw it in your response (but maybe if you include it in your original topic, it would be helpful for others to quickly find as well). Thanks!

Thank you!

10

Thank you Al Gassim. You are correct, metadata automation is the core feature, but not the only useful feature. It can be used also to understand charts and maps, you can just click on any chart and ask it to give you insights. Another very useful area is understanding programs, if you have for example a complex tracker program, this makes it very easy to understand the structure, like the data elements in each stage, what program rules are used in each stage, and to what data elements they relate, and more. You can just navigate to any stage and ask it to check all the program rules in it and fix if any have issues, these are just examples of what can be done.

11

That’s actually really nice! I didn’t know that was possible in the app. :clap: This means it can also be used to help trainees during training!

Since this is LLM, I assume it can also explain using different languages too!

I’m looking forward to how it’ll continue to grow because I see it has the potential to open the door for more proper use of AI within the DHIS2 ecosystem. :+1: