Sekka - a TUI chatbot client
A mornings fun with personal app building and energy tracking
September 24, 2026
So I set out to build my own TUI (Terminal User Interface) LLM client. While other TUI clients exist, none quite did what I wanted, and as much as I love web clients, sometimes I want to stay in a terminal. So we (Qwen 3.8 Flash Next) and I built Sekka
The initial working version took about 20 minutes, and another 20 minutes to setup the GitHub repo nicely with a hosted website, SSL, DNS, etc to land. All in that used 0.04 kWh of relatively clean electricity across 2X DGX Sparks, measured using Spark Dash From an environmental perspective this is pretty great, and efficiency only gets better if you pack on the concurrent work, though this was a single token stream averaging between 40 and 75 tokens/second.
Back to Sekka. Aside from being a TUI app, the neat feature is that the interface offers "knowledge" in the form of tools, that can be configured from the UI or config file. This could let a chat bot answer questions about workflow things "How do I handle credit card payments?" "What are the insurance rules for Texas?" by bounding the responses with knowledge in the file. A good description for the knowledge file will result in accurate responses even with a small model backing it.
The other use case is gaming, like RPGs, where knowledge files could represent character cards, locations, items, monsters, etc. Theres also some light session management to save and resume a chat, as well as context window management which may need some tweaking after I use the tool more.
None of this is super novel, but it fills a need I had, and maybe a need for someone else too. AI has unlocked the ability to build and share these personal apps quickly, and without incredible energy use or environmental impact. What this means for the "value of software" is of course super complicated and something that will take years to shake out.