Building AI That Learns Your Preferences

1

Knowledge Rubrics: Organizing Your World

Every piece of information about you is organized into "knowledge rubrics" — thematic categories that make sense for your life. Each rubric has a name, description, and tracks which sources contributed to it.

class KnowledgeRubric:
    id: int
    knowledge_base_id: int
    name: str  # e.g., "Favorite Cuisines", "Date Night Spots"
    description: str
    source: str  # JSON: ["gmail", "chat", "manual_entry"]
    is_starred: bool
    last_entry_updated_at: datetime

Your Foodie Guru might have rubrics like "Favorite Cuisines," "Dietary Restrictions," "Date Night Restaurants," or "Dishes to Avoid." Your History Keeper organizes memories by "Travel Experiences," "Work Projects," or "Family Milestones."

2

Knowledge Entries: Capturing Real Experiences

Within each rubric, individual knowledge entries capture specific experiences, preferences, or insights. Every entry includes rich metadata about its origin and can connect to multiple rubrics.

class KnowledgeEntry:
    id: int
    knowledge_base_id: int
    title: str
    content: str  # Structured JSON
    source: str  # JSON: ["gmail", "receipt_photo"]
    created_at: datetime
    updated_at: datetime
    rubrics: List[KnowledgeRubric]  # Many-to-many
    attachments: List[KnowledgeEntryAttachment]
    references: List[KnowledgeEntryReference]

3

Attachments & References: Rich Context

Knowledge entries can have attachments (receipt images, photos, documents) and references to other entries or external sources, creating a rich web of interconnected memories.