Her Little Day — A Private AI Companion for Someone I Love

PRATIK KUMAR SINGH — Sun, 04 Oct 2026 18:32:07 +0000

Her Little Day is a private AI-powered daily companion I built for someone I love.
I wanted to solve a simple but real problem: everyday tasks can become overwhelming when they are scattered across a checklist, habits, memories, and personal routines. Instead of building another generic productivity app, I wanted to make something that actually understands the person using it.
Her Little Day combines a personal daily planner with an open-source AI companion that runs locally through Ollama. The companion can understand the user's current day, tasks, habits, personal memories, and context, then provide personalized guidance.
It can also safely propose actions such as creating tasks, completing tasks, rescheduling tasks, or creating habits. Actions require confirmation rather than allowing the AI to directly change important data.
Privacy was a major design goal. The AI inference can run locally on the user's device, while sensitive personal information is protected through strict data isolation and access controls.
I built it specifically for a loved one rather than for a generic audience, so the experience is intentionally personal, gentle, and relationship-focused.

Demo Link:https://drive.google.com/file/d/1eUsWRrrXW551Fk4SYLUAR2JcPUzkDfGs/view?usp=drive_link

Github repo:https://github.com/Code-with-pratik-07/HER-SIDE

How I built it :
I built Her Little Day as a React + TypeScript application with a local-first architecture.
The AI layer uses Ollama with an open-weight language model, allowing the companion to run locally rather than requiring a proprietary AI API.
The companion is built around several layers:
Personal Memory → Context Builder → Local LLM → Response Parser → Safe Action Resolver → Confirmation → Action Executor
The context builder gathers only the information relevant to the current interaction, such as today's tasks, habits, memories, and upcoming events.
I deliberately separated AI-generated suggestions from application actions. The model cannot directly execute arbitrary database operations. Supported actions are validated, checked for ownership and ambiguity, and require user confirmation before execution.
I also added tests around memory isolation, context construction, AI responses, action validation, privacy boundaries, and user-specific data isolation.
The project currently has hundreds of automated tests covering the core application and AI functionality.

Frontend: React, TypeScript, Vite
Styling: Tailwind CSS
State: Zustand
Animation: Framer Motion
Backend/Data: Supabase + PostgreSQL
AI: Ollama + open-weight LLM
PWA/Mobile: Vite PWA / Capacitor
Testing: Vitest

Why Open Innovation Matters:
I could have connected Her Little Day to a proprietary AI API and treated the model as a black box. Instead, I wanted to explore what a personal AI companion could look like when the intelligence is open and locally controllable.
Using an open-weight model through Ollama means the AI can run on the user's own hardware. This changes the privacy model of the application: personal memories, routines, and conversations do not have to be sent to a third-party AI service just to receive assistance.
Open innovation also makes experimentation more accessible. I can change the model, prompts, context strategy, safety layer, and AI behavior without rebuilding the entire product around one proprietary provider.
For something as personal as a companion that understands someone's routines and memories, having control over where inference happens matters to me.
This project showed me that open AI isn't only about having access to a model. It is also about giving developers and users more control over how AI becomes part of their lives.

My Agent Session:
I used an AI coding agent throughout development to help implement, test, debug, and review different parts of the application. I treated the agent as a development collaborator rather than allowing it to make unrestricted product decisions.
The most important engineering decisions — especially privacy boundaries, memory isolation, confirmation-based actions, and the overall companion experience — were intentionally designed and reviewed as part of the project.

Your strongest submission positioning:
I built a private AI companion for someone I love — one that understands her day without needing to send her personal life to a proprietary AI service.

Her Little Day is a private, local-AI companion that turns someone's everyday tasks, habits, and memories into gentle, personalized help

DEV Community: PRATIK KUMAR SINGH

Her Little Day — A Private AI Companion for Someone I Love