Seasoned
The Pitch:
Seasoned is a high-performance, private digital cookbook that bridges the gap between web discovery and kitchen execution. By combining the multimodal power of Gemini 1.5 Flash with a secure, self-hosted PostgreSQL backbone, Seasoned allows users to instantly "distill" messy recipe blogs and food photos into a standardized, searchable, and shareable library they truly own.
Target Audience:
The Modern Minimalist: Cooks who want an ad-free, recipe experience.
The Legacy Keeper: Families digitizing handwritten recipes into a clean digital format.
The Privacy Enthusiast: Users who want the power of AI without storing their personal data in a massive cloud.
The Hybrid Tech Stack:
| Components | Technology |
|---|---|
| Hosting | Private Server (Dockerized on home hardware) |
| CI/CD | Jenkins server |
| Frontend | Nuxt 4 + Vuetify + CSS |
| Backend | Nuxt Nitro |
| Database | Postgres + pgvector |
| Intelligence | Gemini 2.5 Flash |
| Storage | Local File System |
Technical Requirements:
-
AI & Multimodal Intelligence
Multimodal Extraction: Use Gemini 1.5 Flash to accept image/jpeg inputs and return a strictly validated JSON Schema containing title, ingredients, and steps.
Semantic Search: Implement pgvector in the local database. Recipes will be converted into "embeddings" (via Gemini) to allow users to search for "Comfort food for a rainy day" instead of just keyword matches.
-
Full-Stack Architecture (Nuxt 4)
Directory Structure: Adherence to the new app/ directory standard for better IDE performance and separation of concerns.
Serverless-Style Routes: Use Nitro server routes to keep the Gemini API Key hidden from the client-side.
Responsive Design: A UI that adapts perfectly to a tablet propped up on a kitchen counter.
-
Data & Storage
Relational Schema: A PostgreSQL database to manage Users, Recipes, Tags, and Shares.
Private Media Pipeline: A custom upload handler that saves images to a local Docker volume, served via a secured static asset route.
Sharing Permissions: A relational join-table logic that allows one user to "push" a recipe to another user's library.
Use Cases:
Photo-to-Recipe: User snaps a picture of a magazine page; Gemini extracts the text; the user saves it to their Postgres DB.
Semantic Discovery: User searches for "High protein dinner with lime" and the app uses vector similarity to find the best match.
Ad-Free Web Scraping: User pastes a blog URL; the server fetches the content, and Gemini strips out the ads and life stories.
Collaborative Boxes: One user "seasons" a recipe (rates/tags it) and shares it with someone who also uses the instance.