Knowledge Store
The Knowledge Store is Julia's structured data layer — a persistent, organized database where she stores and retrieves domain-specific information across your conversations. Think of it as Julia's notebook, organized by topic.
How It Works
Data in the Knowledge Store is organized into namespaces (collections) and keys (entries). Julia uses this automatically when tasks require persistent, structured data.
Data Structure
namespace: "apartment_search"
├── key: "criteria" → { budget: 1500, rooms: 2, city: "Paris" }
├── key: "shown_listings" → [{ id: "apt-1", ... }, { id: "apt-2", ... }]
└── key: "favorites" → [{ id: "apt-3", ... }]
namespace: "research"
├── key: "quantum_computing_notes" → { ... }
└── key: "reading_list" → [{ ... }]Knowledge Store vs Memory
Knowledge Store
- Structured JSON data (objects, arrays, numbers)
- Organized by namespace and key
- Best for: project tracking, criteria, lists, state
- Loaded on demand when relevant
Memory
- Free-text observations and facts
- Retrieved via semantic similarity search
- Best for: preferences, facts, past events
- Automatically included in context
Examples
Apartment search tracking
Julia
Online
I'm looking for a 2-bedroom apartment in Paris, max €1,500/month, near a metro station.
I've saved your search criteria: 🏠 **Apartment Search** • Rooms: 2 bedrooms • Budget: ≤€1,500/month • Location: Paris • Requirement: Near a metro station I'll use these criteria going forward. Want me to set up a daily job to check listings for you?
Research project notes
Julia
Online
Save these research findings under my quantum computing project: the new IBM paper shows 1,000+ qubit processors are viable by 2027.
Saved to your research notes: 📝 **quantum_computing → findings** • IBM paper: 1,000+ qubit processors viable by 2027 • Added to existing findings (now 4 entries) I can retrieve these anytime you need them.
Operations
get / setRead or write a value by namespace + key
appendAdd an item to an array
listList keys in a namespace or all namespaces
deleteRemove an entry or item from an array
Works with Scheduled Jobs
Scheduled jobs can automatically load knowledge store data at runtime. For example, an apartment search job loads your criteria and previously shown listings, ensuring it only reports new matches. This is configured via dataNamespaces when creating a job.
Availability
The Knowledge Store is available on all plans. Julia uses it automatically when tasks involve tracking structured data over time. You can also view your stored data from the web portal.