HomeMemory {Memo}

Description and use

HomeMemory is a personal memory assistant that lets users photograph items, add notes, and query them in natural language using a local Gemma 4 model and MongoDB Atlas for storage and retrieval. It reduces search time and improves reliability, especially for family users.

Problem

Families forget small but important items and dates, such as where a spare key is or when a medicine expires.

Outcome

The system achieved a top‑1 retrieval accuracy of 78% on 10 real items, reducing retrieval time from 10‑15 minutes to 2‑3 minutes.

Workflow

• Take or upload a photo of an item
• Add a short note about its location
• Gemma 4 (via Ollama) processes the photo and outputs structured JSON
• EmbeddingGemma creates a vector for the memory
• MongoDB Atlas Vector Search stores and retrieves memories
• User queries in English/Urdu; system returns the photo and note
• Optional confirm‑and‑edit before saving
• Threshold controls whether a match is returned or "not found"

Evidence

Gemma 4 runs locally through Ollama. It reads each photo and returns structured JSON (category, item, description, readable text, expiry date) and writes the final answers.

Source

MongoDB Atlas Vector Search stores the memories and finds the closest match to a question.

Source

top-1 retrieval accuracy was [X/Y = 78%] on 10 real items.

Source

I built HomeMemory by working with Claude in a chat window, then running and testing everything on my own laptop.

Source