Ghana sits under some of the strongest sunshine on earth. Solar supplies
about 1–3% of our grid electricity. The gap isn't desire — it's guesswork.
Families want solar, but nobody can tell them plainly: what size? what cost?
will it pay for itself?
So I built the answer. On a phone. Between classes.
What it does
A Streamlit app that lets any household:
- Enter their real appliances (fridge, TV, fans, pump…)
- Get a sized system: panels (kW), battery (kWh), inverter (kW), using load math + region-aware peak sun hours
- See an estimated cost in cedis
- See the payback computed from their real ECG bill (bill ÷ kWh = true tariff — no invented numbers)
- Request a professional quote → the sized lead is emailed to me
🔗 Live app: https://ghana-solar-tool-pli6auappdapghergd2rfz.streamlit.app
📂 Code: https://github.com/collins7-code/ghana-solar-tool
The bugs (an honest post-mortem, because that's the real lesson)
1. The form that crashed on submit.
I pasted my lead-capture form above the sizing code. It rendered fine
(empty boxes need no math) — and threw NameError: daily_kwh is not defined
the moment someone submitted. Lesson: test the submit path, not just the
render path.
2. "Make sure you open this page through a web server."
FormSubmit expects browser calls with Origin/Referer headers. My app
posts server-to-server, so it looked like a local HTML file and got refused.
Fix: attach the live app URL as Origin + Referer headers.
3. The silent activation step.
First submission to a new email only triggers a one-time activation email.
Until you click it, leads are accepted but never delivered. I chased a
"broken" pipeline that was just waiting for one click.
What real users taught me
My first real user (a friend) left the appliance table at its defaults —
so her quote wasn't truly hers. Real hands > my assumptions. That's now an
open UX issue: nudge users to personalise the table.
What's next — and where you come in
This project is built in the open. Open issues include: real ECG tariff
bands, measured irradiance data, Twi localization, and mobile UX.
I'm also opening an open-data track with real Ghana infrastructure data.
To show you why it matters: in 2010 Ghana spent US$171M on road
maintenance — yet 58% of the 68,053 km network was still rated fair or
poor. Where should the next cedi go? That's the kind of question this
track will explore, roads first, then energy.
Grab a good first issue. Read CONTRIBUTING.md. Come build with me.
One line for the road
I lost version 1 of this project when my phone broke. I rebuilt it better.
The builder survived. ⚡
Top comments (3)
Really impressive that you built this on a phone, especially at 17. The “real hands > assumptions” lesson is spot on — testing with actual users always exposes UX issues you don’t see yourself.
Thanks for that, I really needed it
렌더링 경로는 통과했지만 제출 순간에 daily_kwh가 정의되지 않아 깨졌다는 부분이 특히 좋았습니다. 입력값을 비워둔 초기 화면만 보는 테스트와 실제 제출 경로 테스트를 분리해 둬야 한다는 교훈이 명확하네요. 기본 appliance 값 때문에 첫 사용자의 견적이 개인화되지 않았다는 관찰도 중요해서, 결과 계산 전에 사용자가 최소 한 항목을 수정했는지 확인하는 단계가 도움이 될 것 같습니다.