Beyond Exact Keyword Search
Traditional journals require you to remember the exact keyword you typed years ago. If you wrote “strolled along the canal”, a search for “river walk” would return zero results.
Thenvue uses semantic vector search to understand the meaning behind your memories. You can search conceptually:
- “Books I enjoyed reading last autumn”
- “Conversations about moving to a new apartment”
- “Restaurants I visited with Maya in Mumbai”
- “How I felt before starting my new project”
How the “Ask” Feature Works
When you submit a question in Thenvue's Ask tab:
- Semantic embedding: Your query is converted into a vector representation.
- Database retrieval: Thenvue searches your private memory records in PostgreSQL using cosine distance to identify relevant moments.
- Grounded summary: The most relevant memories are passed as context to synthesize a concise, factual summary of what happened.
- Source transparency: The underlying memory cards are provided so you can click through to read your original words and see any attached photos.
Real Capabilities and Clear Boundaries
We believe in honest technology. Here is what AI memory search does and does not do:
Retrieves and summarizes memories you have actually recorded and saved in your space.
It cannot recall moments you never recorded, guess details outside your notes, or provide perfect photographic recall.
Your Data & Privacy
Memory embeddings and queries are scoped strictly to your authenticated account using database Row-Level Security. We do not use your private personal journal entries to train public foundational AI models. For in-depth technical details, visit our AI & Data Transparency page.