What it is
Omakase is my anime recommendation counter. Instead of another endless catalog, it makes a small menu from the shows I have scored and the qualities I actually enjoy. Each recommendation comes with a reason to try it.
It remains part of my everyday toolkit. The public counter lets visitors bring their own model key, a public AniList username or a MyAnimeList export, and optional taste notes.
Bring your own taste
- Choose a supported model provider and supply your own key. Provider usage is billed to your provider account.
- Bring scored watch history from AniList or a MyAnimeList export.
- Add what matters to you, then read a focused menu with explanations.
Guest requests use request-local credentials and taste notes. The hosted service receives those inputs to prepare the recommendation and call the selected provider. This is separate from saving a profile in a Lite account.
The illustrated study explains a recommendation with fictional examples before a visitor submits anything. Returned menus keep the full reason and history pairing, label model estimates, and let a visitor compare two picks on the same page.
A counter I can come back to
Lite accounts are invitation-only. An invited account can retain a profile, an encrypted provider key, and recommendation feedback such as Already Watched. That persistent account experience has a different storage boundary from a guest request.
Omakase recommends anime. It does not expose the private library automation or playback system behind Omakase Plus.
The engineering behind the menu
I use Python and FastAPI to connect watch-history inputs with a model provider and turn the response into a focused recommendation experience. My role spans the product flow, API integration, agent-assisted implementation, and verification of the hosted result.
A key design decision is separating guest requests from saved accounts. A guest supplies inputs for one request; an invited Lite account has a different persistence boundary. The interface needs to make that distinction understandable before someone submits a key.
Why it stays in the workshop
The useful part is the conversation between taste and history: which patterns to repeat, which familiar recommendations to avoid, and what might be worth trying next. I still use it, so it belongs among my current tools.
Explore Omakase Plus for the separate private library companion.