-
-
Owner's compact Week in Review on desktop.
-
The same Owner review at mobile width
-
Partner's separate private view with the same shared household content
-
Deterministic weekly status and no more than three priorities
-
A grounded finance observation with evidence details.
-
Private health comparison and a unit-correction notice without medical advice
-
The shared household overview and quick capture entry point
Busy couples keep money, health records, household plans, and upcoming obligations across spreadsheets, apps, and messages. Mithra, meaning friendship, brings those into one shared view while letting each adult keep selected records private.
Users can capture an update in their own words or import CSV, XLSX, and PDF files. AI is used to categorize the updates into Finance, Health, and Planning views. Mithra shows trends, offers insights, and events, working as a coach, without offering any diagnosis.
The core experience is Week in Review. It starts with a clear household status, ranks no more than three priorities, groups only clearly related planning and finance records, shows what changed and what is next, and gives a factual observation from valid data.
An optional model provider can improve the wording of evidence-linked coaching, but it cannot invent facts, change a record, spend money, make a booking, provide medical or financial advice, or mediate a relationship.
Mithra is a low-dependency Go application with an embedded UI, SQLite, encrypted sources and backups. It is designed to be self-hosted in lightweight VPS and users bring their own API key or can use local LLMs via Ollama.
How we built it
Mithra uses Go, SQLite, embedded HTML/CSS/JavaScript, and a single-binary deployment model.
GPT 5.6 was used end-to-end, from planning using Sol, to implementation using Terra and Luna subagents, with Sol acting as the verifier. I was responsible for the vision and critique, and GPT the executor and thought partner
What is next
Onboard a new real-world users, validate, and refine on the app with their feedback before widening the scope for Mithra. We will learn from real household use before adding broader family roles or more domains. The next work will focus on better import correction, stronger trend views, and carefully bounded coaching while keeping people in control.
Built With
- codex
- css
- go
- gpt-5.6
- html
- javascript
- openai
- remotion
- sqlite