GitHub Flows
Controlled workflows triggered by GitHub events. Agents, checks, and publication steps become one observable process instead of a hidden automation chain.
Independent systems engineer · Latvia
Controlled systems for work with AI agents.
I build practical infrastructure where intent becomes bounded agent work and every result stays observable. Start with GitHub Flows on a real repository, then decide how far to take it.
- 25+ years in web engineering
- JavaScript and Node.js
- Self-hosted by design
- Multilingual delivery
GitHub Flows / controlled path
01Intent and boundary
02GitHub event
03Bounded agent action
04Observable result
event → agent reaction → new observable event
Who I am
Senior engineering without a black box.
I am an independent web developer and systems engineer. I work at the intersection of JavaScript architecture, self-managed infrastructure, and agent-driven development.
My products are also my working tools: I use TeqFW to build them, ADSM to maintain their context, and GitHub Flows to make agent activity bounded and visible.
What I build
One engineering base, three practical products.
Each product solves a different part of the same problem: keeping digital work structured while agents take on more of the execution.
Tequila Framework
My modular JavaScript platform and architectural foundation for systems that must remain legible, replaceable, and under control as they grow.
AI-generated multilingual sites
Fast, lightweight content sites on TeqFW and TeqCMS, with pages, translations, and publication maintained through an agent-driven workflow.
How I work
Agents need context, limits, and evidence.
ADSM is my method for keeping product meaning, architecture, constraints, and verification available to both people and agents. It turns agent-driven work into a managed engineering process.
The book Agent-Driven Development explains the method; the projects and publications show it in practice.
Read the English edition Browse projects Explore the library
- Intent remains with the person.
- Context survives beyond one chat.
- Authority is explicit and bounded.
- Results are checked and observable.
A small first step
Try one controlled agent workflow on your repository.
Up to five test tasks, visible GitHub results, execution evidence, and a concise engineering recommendation.