# Virtual Software Office

> A live, self-running simulation of a fictional Singapore software company ("Pranav Soft Solutions"), viewable at https://vo.ishaugur.com/. Created by Ravi B, AI Engineer who turns AI agents and workflow automation into systems that run on their own.

This is a simulation: all people, companies, clients, projects and money are fictional.

## What it is
- A virtual software office that runs itself on Singapore working days (09:00–18:00 SGT), advancing every 15 minutes.
- Visitors watch a live floor plan: staff in cubicles, cabins and meeting rooms, working from home, taking breaks, with AI-written chat bubbles.
- Dashboards show projects (progress, man-days, risks), hiring (pipeline, candidates), people (teams, leave balances) and an office log.

## How it works
- Rules and data: PostgreSQL functions (attendance, meetings, project stages, hiring, leave approvals, payroll, CPF, corporate tax).
- Orchestration: n8n workflows (clock, AI chatter, Telegram commands, web data, alerts).
- AI: Groq-hosted open models (gpt-oss-20b / gpt-oss-120b) write staff chat, the manager's notes, lead reviews and answer the owner's questions – always from database facts.
- Control: the owner steers the company from Telegram (leads, hiring, appraisals, ad-hoc events, questions).
- Web: static page + cached live data behind nginx and Traefik.

## Key pages
- [Live office](https://vo.ishaugur.com/): the running simulation
- [About](https://vo.ishaugur.com/about): purpose, stack and creator
- [FAQ](https://vo.ishaugur.com/faq): common questions

## Facts
- Live since 21.10.2026. Last updated 22.10.2026.
- Free to view. No sign-up. Feedback form on the live page.
- Creator: Ravi B, AI Engineer (AI agents, workflow automation, n8n, PostgreSQL, LLM integration).

## How it was built
- Semi-automated development: designed and coded with the AI agent Claude Opus 5.5 (Anthropic) via Claude Code, architected, reviewed, deployed and tested by Ravi B.
- Built in about 3 calendar days (20-22.10.2026); estimated 30-45 working days if developed manually.
- About 227 million tokens processed (0.93M output, 2.05M new input, 224M cached input), 788 model calls.
- 71 SQL files (~11,400 lines), 156 database functions, 45+ tables, 12 n8n workflow files.