TAMSIV is in production — from idea to Play Store in 6 months

Six months ago, there was a piece of paper stuck to my fridge. On it, three words scrawled in marker: "voice task app". Today, TAMSIV is in production on the Google Play Store. In between, 750+ commits, 24 features, 6 languages, and a solo dev who didn't get much sleep.

This article is the story of those six months. Not a polished, corporate post-mortem. The real journey, with the doubts, the pivots, the nights debugging a voice pipeline at 3 AM, and that brutal satisfaction when Google finally approves your production build.

Key points

• 6 months of solo development, from October 2025 to April 2026, with 750+ commits on a frontend + backend + website monorepo.
• 24 features delivered: AI voice assistant, calendar, gamification, collaborative groups, image generation, i18n 6 languages, and more.
• 60% of web traffic comes from non-French-speaking visitors thanks to internationalization in 6 languages.
• TAMSIV has been available for free on the Google Play Store since April 4, 2026.

October 2025: the piece of paper on the fridge

It all started with a common frustration. Grocery lists on scraps of paper. Diving club tasks scattered across WhatsApp. Forgotten reminders in an app no one ever opened. Information was systematically lost between different platforms, people, and times of day. I had mentioned it in the very first blog post.

The idea was simple: what if you could manage everything by voice? Not an assistant that just transcribes what you say, but an AI that understands context, knows that "Tuesday's meeting" means next Tuesday, detects that a task is urgent without you having to specify it, and organizes everything for you while you're driving or doing the dishes.

In October 2025, I opened a terminal and typed npx react-native init. The first commit dates back to October 3rd. Six months later, this project is in production.

24 features: what TAMSIV contains today

Here's what was built in 6 months, feature by feature. Each item on this list represents days of work, architectural choices, and sometimes complete refactorings.

The core: the AI voice assistant

The voice recorder is TAMSIV's entry point. You speak, the AI understands, and it creates tasks, memos, or calendar events. The voice pipeline goes through the phone's native speech recognition, a WebSocket backend, an LLM via OpenRouter, and audio feedback via OpenAI TTS. I detailed this entire chain in the article on the voice pipeline.

The AI doesn't just transcribe. It calls functions: create_task, update_task, create_memo, create_calendar_event. It detects your naming patterns, suggests priorities, and understands complex instructions like "move Tuesday's meeting to Thursday and notify the group."

The 24 features

1. AI Voice Assistant — Voice recorder with real-time STT + LLM + TTS pipeline
2. Task Management — Priorities, deadlines, recurrence, subtasks, favorites
3. Rich Memos — Text, images, rich text editor with formatting
4. 4-View Calendar — Day, week, month, year on mobile and web
5. Collaborative Groups — 6-level hierarchy, roles, permissions
6. Collaborative Checklists — Individual or collective validation
7. Task Assignment — Within groups, with progress tracking
8. AI Image Generation — Via Runware HiDream or Gemini, directly from the voice recorder
9. Gamification System — 12 levels, 10 badges, streaks up to 365 days, daily challenges
10. Activity Feed — Unified timeline of all actions
11. Notifications and Reminders — FCM push, recurring reminders, local notifications
12. Contextual Search — Full-text search on tasks, memos, events
13. Referral System — Invitation codes with rewards (free Pro month)
14. Subscriptions — Free, Pro, Team via RevenueCat with feature gates
15. Attachments — Photos, videos, documents on tasks and memos
16. QR Code Authentication — Password-less mobile/web login
17. Voice Customization — Configurable AI voice (OpenAI TTS)
18. Web Dashboard — Next.js with near-complete parity with mobile
19. Admin Panel — User stats, metrics, Recharts graphs, configuration
20. Internationalization — 6 languages (FR, EN, DE, ES, IT, PT) across the entire stack
21. Monitoring — Crashlytics (frontend) + Sentry (backend) + email alerts
22. Rich Text Editor — Formatting, lists, checklists in memos
23. Drag and Drop — Reordering of elements by dragging and dropping
24. STT Dual Mode — Native (device) speech recognition with Deepgram cloud fallback

Technical challenges that marked these 6 months

The real-time voice pipeline

The biggest technical challenge of the project. Transmitting audio from a phone microphone, converting it to text, sending it to an LLM, executing actions, generating a voice response, and playing it — all in under 3 seconds. Bidirectional WebSocket, timeout management, fallback between native STT and Deepgram cloud, audio cleanup with a 30-second security timeout. I compared the approaches in the article on native STT vs Deepgram.

The 3-schema database

From the start, I structured Supabase into three PostgreSQL schemas: privat for user data, collaborative for groups, and gamification for stats and badges. This choice, documented in the article on DB restructuring, avoided a lot of pain later on. RLS (Row Level Security) policies are granular: each table has its own access policies.

End-to-end internationalization

Internationalizing an app isn't just translating strings. It's translating the mobile frontend (React Native), the backend (error messages, AI responses), the website (Next.js with next-intl and localized routing), Play Store listings, and even blog URL slugs. The result: 6 languages, 60% non-French-speaking web traffic. Germany is the 3rd market without a single post in German. I detailed this strategy in the article on i18n as an acquisition channel.

React Native New Architecture

TAMSIV runs on React Native 0.81 with the New Architecture (Fabric). The transition was not without pain: some libraries did not yet support Fabric, requiring shims, patches, and a lot of patience. But the result is there: fluid 60fps animations, native transitions, and a startup time halved compared to the old architecture.

The two-level cache

To ensure the app is responsive even offline, I implemented a ContentCacheService with two levels: memory cache (JavaScript Map) for instant access, and AsyncStorage for persistence. All synchronized in real-time via Supabase Realtime. When you open the app, data displays immediately from the cache and updates silently in the background.

6 languages, 60% non-French traffic

The number that surprised me the most. In March 2026, when I launched automatic translation into 6 languages (French, English, German, Spanish, Italian, Portuguese), 60% of incoming traffic to tamsiv.com came from non-French-speaking visitors. Germany, Spain, Brazil — markets I hadn't targeted at all.

I18n became my primary free acquisition channel. Every blog post translated into 5 additional languages means 5 more entry points into Google. The Play Store is also localized in 6 languages: title, short and long description, release notes. For a solo dev with no marketing budget, this is a disproportionate lever.

The road to production

Going into production isn't a single moment. It's a process. Here are the steps that marked the last few months:

• February 2026 — First submission to the Play Store, first rejection (metadata policy). Correction and resubmission.
• March 2026 — 12 alpha testers recruited. Intensive 14-day testing phase. Zero crashes reported.
• End of March 2026 — Zero feature sprint: 30+ polish commits, no new features. Crashlytics + Sentry deployed. I talked about it in the article on the final sprint.
• April 2, 2026 — Google approves the production build.
• April 4, 2026 — TAMSIV is publicly available on the Play Store.

750+ commits. 42 blog posts (including this one). A backend deployed on Railway. A website on Vercel. A Supabase database with hundreds of RLS policies. And an Android app that does exactly what the piece of paper on the fridge promised: manage your tasks by voice, simply.

Download TAMSIV

TAMSIV is available for free on the Google Play Store. The Free plan includes the AI voice assistant, task and memo management, calendar, and notifications. Pro and Team plans unlock collaborative groups, AI image generation, and advanced gamification.

Download TAMSIV on the Google Play Store

You can also discover the web dashboard on tamsiv.com to manage your tasks from your computer.

Frequently Asked Questions

Is TAMSIV free?

Yes. The Free plan provides access to the AI voice assistant, task and memo management, calendar, and reminders. Pro (advanced features) and Team (group collaboration) plans are available as monthly or annual subscriptions via the Play Store.

On which devices does TAMSIV work?

TAMSIV is available on Android (minimum Android 7.0 / API 24). The iOS version is planned for later. A web dashboard is accessible at tamsiv.com/app to manage your tasks from a computer.

How does the voice assistant work?

You speak into the voice recorder, your phone's native speech recognition transcribes your voice into text, an LLM interprets your request and executes actions (create a task, memo, event), then a synthetic voice confirms what has been done. All in real-time via WebSocket.

Is my data secure?

Data is stored on Supabase (PostgreSQL) with granular Row Level Security (RLS) policies. Authentication goes through Supabase Auth (email, magic link, QR code). Communications are encrypted with HTTPS/WSS. No voice data is stored: audio is processed in memory and deleted immediately after transcription.

Is TAMSIV available in multiple languages?

Yes, in 6 languages: French, English, German, Spanish, Italian, and Portuguese. The interface, website, Play Store, and even blog articles are translated. The voice assistant understands and responds in the configured language.

How long did development take?

6 months of solo development (October 2025 - April 2026), 750+ commits, 24 major features. The project is developed "build in public" with a technical blog documenting each step from day one.