Overview
Labels are CORE’s flexible tagging system for organizing episodes, sessions, and documents within your workspace. Think of them as project tags or categories that help you filter and find related memories quickly. Unlike traditional hierarchical folders, labels use a flat, Linear-style organization that lets you apply multiple tags to any piece of content, making it easy to organize information across different contexts and projects.What Are Labels?
Labels are workspace-unique tags that you can assign to:- Episodes: Individual conversations and interactions
- Sessions: Related groups of conversations
- Documents: Files and content you’ve ingested
- Name: A unique identifier within your workspace (max 100 characters)
- Description: Optional details about what the label represents
- Color: Visual identifier for quick recognition
- Workspace Scope: Labels are unique per workspace, not global
Key Characteristics
Workspace-Scoped
Labels are scoped to your workspace, meaning:- Each workspace has its own set of labels
- Label names must be unique within a workspace
- Different workspaces can have labels with the same name
- You can’t share labels across workspaces
Multiple Labels per Item
You can assign multiple labels to any episode:- Episodes have a
labelIdsarray field - No limit on the number of labels per episode
- Filter searches by one or multiple labels
- Labels don’t nest or form hierarchies
How Episodes Use Labels
When you ingest content or have conversations, episodes can be tagged with labels:Creating and Managing Labels
Via Web Interface
- Navigate to Settings → Labels
- Click “Create Label”
- Enter name, optional description, and choose a color
- Labels appear in your workspace immediately
Via API
List all labels in your workspace:Using Labels in Search
Labels provide powerful filtering capabilities across all search methods:Basic Search with Labels
Multiple Label Filtering
Search across multiple labels to find related content:Recent Work by Label
Combine labels with time filters:Label-Based Search Architecture
When you filter by labels, CORE’s search system:- BM25 Search: Filters episodes by
labelIdsbefore scoring - Vector Search: Applies label filter to episode embeddings
- Episode Graph Search: Constrains graph traversal to labeled episodes
- BFS Search: Limits entity exploration to episodes with matching labels
API Endpoints for Label Management
| Endpoint | Method | Description |
|---|---|---|
/api/v1/labels | GET | List all labels in workspace |
MCP Integration with Labels
CORE’s MCP server exposes memory tools that support label filtering:Available Tools
memory_search: Search with optionallabelIdsarraymemory_ingest: Store conversations withlabelIdsmemory_get_labels: List all workspace labelsmemory_get_document: Retrieve specific documents
Best Practices
Label Naming
- Use clear, descriptive names (e.g., “TaskMaster Project”, “Authentication”)
- Keep names concise but meaningful
- Use consistent naming conventions across your workspace
Label Organization
- Create labels for projects, topics, or contexts
- Don’t over-label - use 2-4 labels per episode typically
- Use favorites for truly important items rather than creating “Important” labels
