Zgidno-Vidpovidno
Zgidno-Vidpovidno is an AI-powered Ukrainian report generator built with Astro, React, Firebase Hosting, and a Cloudflare-proxied LLM backend.
Categories:
AIWebAstroReactTypeScriptFirebaseCloudflareOpen SourceGitHubSocial Impact
Zgidno-Vidpovidno is a bilingual Ukrainian AI web app that turns everyday civilian incidents into absurdly formal military-style reports. It demonstrates Vasilkoff Ltd's practical AI product delivery: prompt design for Ukrainian-language generation, structured JSON output, React UX, Cloudflare-proxied model access, Firebase Hosting, analytics, and rate-limited public usage.
Overview
Zgidno-Vidpovidno is a small but complete AI product that converts ordinary Ukrainian phrases into formal, over-engineered military-style reports. The app preserves the practical meaning of the user's incident while rewriting it in a deliberately bureaucratic tone with report text, resolution, order, approvers, regulation, authorization, and operation code.
The project is intentionally playful, but the implementation demonstrates a serious workflow: structured LLM output, prompt constraints, Ukrainian-language few-shot examples, public-facing rate limiting, analytics, and a responsive React interface.
What It Demonstrates
- Ukrainian-language prompt engineering with stable JSON output.
- Astro and React delivery for a fast static frontend.
- Cloudflare Worker proxying so model access is not exposed directly in the browser.
- Firebase Hosting and Firebase Analytics integration.
- Practical UI actions including copy, share, branch selection, length controls, and generation status.
- Open-source delivery with a concise README and reproducible build.
Architecture
The frontend is built with Astro, React, and TypeScript. Users select a service branch, choose generation length, enter a short incident, and submit it to a Cloudflare-proxied LLM endpoint. The model is instructed to return a strict JSON object that the UI validates and renders into a report-like document.
This architecture keeps the app lightweight while still showing the same concerns that matter in client AI products: prompt reliability, response parsing, UX feedback, quota protection, analytics events, and deployment discipline.
Why This Case Matters
Zgidno-Vidpovidno is a compact example of how Vasilkoff Ltd can turn a culturally specific idea into a working AI web product quickly. It is especially relevant for teams that need local-language AI tooling, structured generation, internal workflow automation, or public demos that are simple enough to maintain but polished enough to use.
Open the live app or view the source on GitHub.
Project site: zgidno-vidpovidno.web.app