KenyaADK — first ADK implementation for East African civic data (budget, parliament, drought, rights EN/SW) #5484
Replies: 2 comments
|
Following up on the KenyaADK discussion. The East African civic AI portfolio has expanded significantly since that post: New tools built on or complementing ADK:
Portfolio now: 37 repos · 29 tools · 4 PyPI packages · 4 MCP servers The ADK + MCP + A2A stack continues to serve as the infrastructure layer. Would still love to contribute a cookbook entry on civic data agent patterns — especially the bilingual (EN/SW) interaction patterns that work well for East African audiences. |
|
Fantastic real-world use of ADK's sub_agents pattern. Routing drought, budget, parliament and bilingual EN/SW rights queries through one root agent on top of your own MCP and A2A servers is exactly the kind of civic-tech application we love seeing. Closing this out since it's a third-party project rather than an ADK core issue. Did see your offer to write a cookbook entry on the bilingual patterns, genuinely good idea, worth raising directly on adk-docs so it reaches the right people. |
Uh oh!
There was an error while loading. Please reload this page.
What I built
KenyaADK — the first Google ADK implementation for East African civic data. Four specialist agents orchestrated by a root
LlmAgent, all running on Gemini 2.0 Flash free tier.Agents and tools
drought_agentget_drought_status(county)budget_agentget_county_budget(county)parliament_agentquery_parliament(topic)rights_agentget_constitutional_right(topic, lang)The full protocol stack
KenyaADK demonstrates ADK, MCP, and A2A working together:
Run it
Data
Built on Kenya Civic Datasets published to Kaggle, HuggingFace, and Zenodo with academic DOIs — all CC BY-SA 4.0.
Portfolio context
Part of a 24-tool East African civic AI portfolio: gabrielmahia.github.io
Happy to contribute a cookbook entry on civic data RAG + agent patterns for ADK documentation.
All reactions