Spacedrive Semantic Tagging System

Overview

The Spacedrive semantic tagging system is an advanced, graph-based tagging architecture that transforms traditional flat tagging into a sophisticated semantic fabric for content organization. Unlike simple label-based systems, semantic tags support polymorphic naming, context-aware disambiguation, hierarchical relationships, and intelligent conflict resolution during synchronization. This system implements the semantic tagging architecture described in the Spacedrive whitepaper, enabling enterprise-grade knowledge management capabilities while maintaining intuitive user experience.

Core Architecture

Design Principles

Graph-Based DAG Structure - Tags form a directed acyclic graph with closure table optimization
Polymorphic Naming - Multiple tags can share the same name in different contexts
Semantic Variants - Each tag supports formal names, abbreviations, and aliases
Context Resolution - Intelligent disambiguation based on existing tag relationships
Union Merge Conflicts - Sync conflicts resolved by combining tags (additive approach)
AI-Native Integration - Built-in confidence scoring and pattern recognition
Privacy-Aware - Tags support visibility controls and search filtering

Core Components

SemanticTag - Enhanced tag entity with variants and relationships
TagRelationship - Typed relationships between tags (parent/child, synonym, related)
TagClosure - Closure table for efficient hierarchical queries
TagApplication - Context-aware association of tags with content
TagUsagePattern - Co-occurrence tracking for intelligent suggestions
TagContextResolver - Disambiguation engine for ambiguous tag names

Data Models

SemanticTag

The core tag entity with advanced semantic capabilities:

pub struct SemanticTag {
    pub id: Uuid,

    // Core identity
    pub canonical_name: String,           // Primary name (e.g., "JavaScript")
    pub display_name: Option<String>,     // Context-specific display

    // Semantic variants - multiple access points
    pub formal_name: Option<String>,      // "JavaScript Programming Language"
    pub abbreviation: Option<String>,     // "JS"
    pub aliases: Vec<String>,            // ["ECMAScript", "ES"]

    // Context and categorization
    pub namespace: Option<String>,        // "Technology", "Geography", etc.
    pub tag_type: TagType,               // Standard, Organizational, Privacy, System

    // Visual and behavioral properties
    pub color: Option<String>,           // Hex color for UI
    pub icon: Option<String>,            // Icon identifier
    pub description: Option<String>,     // Human-readable description

    // Advanced capabilities
    pub is_organizational_anchor: bool,   // Creates visual hierarchies in UI
    pub privacy_level: PrivacyLevel,     // Normal, Archive, Hidden
    pub search_weight: i32,              // Influence in search results

    // Compositional attributes
    pub attributes: HashMap<String, serde_json::Value>,
    pub composition_rules: Vec<CompositionRule>,

    // Metadata
    pub created_at: DateTime<Utc>,
    pub updated_at: DateTime<Utc>,
    pub created_by_device: Uuid,
}

TagType Enum

pub enum TagType {
    Standard,      // Regular user-created tag
    Organizational,// Creates visual hierarchies in interface
    Privacy,       // Controls visibility and search behavior
    System,        // AI or system-generated tag
}

PrivacyLevel Enum

pub enum PrivacyLevel {
    Normal,   // Standard visibility in all contexts
    Archive,  // Hidden from normal searches but accessible via direct query
    Hidden,   // Completely hidden from standard UI
}

TagRelationship

Defines relationships between tags in the semantic graph:

pub struct TagRelationship {
    pub parent_tag_id: i32,
    pub child_tag_id: i32,
    pub relationship_type: RelationshipType,
    pub strength: f32,              // 0.0-1.0 relationship strength
    pub created_at: DateTime<Utc>,
}

pub enum RelationshipType {
    ParentChild,  // Hierarchical relationship (Technology → Programming)
    Synonym,      // Equivalent meaning (JavaScript ECMAScript)
    Related,      // Semantic relatedness (React Frontend)
}

TagApplication

Context-aware association of tags with user metadata:

pub struct TagApplication {
    pub tag_id: Uuid,
    pub applied_context: Option<String>,    // "image_analysis", "user_input"
    pub applied_variant: Option<String>,    // Which name variant was used
    pub confidence: f32,                    // 0.0-1.0 confidence score
    pub source: TagSource,                  // User, AI, Import, Sync
    pub instance_attributes: HashMap<String, serde_json::Value>,
    pub created_at: DateTime<Utc>,
    pub device_uuid: Uuid,
}

pub enum TagSource {
    User,    // Manually applied by user
    AI,      // Applied by AI analysis with confidence scoring
    Import,  // Imported from external source
    Sync,    // Synchronized from another device
}

Database Schema

Tables Overview

-- Core semantic tags
CREATE TABLE semantic_tags (
    id INTEGER PRIMARY KEY,
    uuid BLOB UNIQUE NOT NULL,
    canonical_name TEXT NOT NULL,
    display_name TEXT,
    formal_name TEXT,
    abbreviation TEXT,
    aliases JSON,                    -- Array of alternative names
    namespace TEXT,                  -- Context grouping
    tag_type TEXT DEFAULT 'standard',
    color TEXT,
    icon TEXT,
    description TEXT,
    is_organizational_anchor BOOLEAN DEFAULT FALSE,
    privacy_level TEXT DEFAULT 'normal',
    search_weight INTEGER DEFAULT 100,
    attributes JSON,                 -- Key-value pairs for complex attributes
    composition_rules JSON,          -- Rules for attribute composition
    created_at TIMESTAMP NOT NULL,
    updated_at TIMESTAMP NOT NULL,
    created_by_device UUID,

    UNIQUE(canonical_name, namespace) -- Allow same name in different contexts
);

-- Hierarchical relationships
CREATE TABLE tag_relationships (
    id INTEGER PRIMARY KEY,
    parent_tag_id INTEGER NOT NULL,
    child_tag_id INTEGER NOT NULL,
    relationship_type TEXT DEFAULT 'parent_child',
    strength REAL DEFAULT 1.0,
    created_at TIMESTAMP NOT NULL,

    FOREIGN KEY (parent_tag_id) REFERENCES semantic_tags(id) ON DELETE CASCADE,
    FOREIGN KEY (child_tag_id) REFERENCES semantic_tags(id) ON DELETE CASCADE,
    UNIQUE(parent_tag_id, child_tag_id, relationship_type)
);

-- Closure table for efficient hierarchy traversal
CREATE TABLE tag_closure (
    ancestor_id INTEGER NOT NULL,
    descendant_id INTEGER NOT NULL,
    depth INTEGER NOT NULL,
    path_strength REAL DEFAULT 1.0,

    PRIMARY KEY (ancestor_id, descendant_id),
    FOREIGN KEY (ancestor_id) REFERENCES semantic_tags(id) ON DELETE CASCADE,
    FOREIGN KEY (descendant_id) REFERENCES semantic_tags(id) ON DELETE CASCADE
);

-- Enhanced tag applications
CREATE TABLE user_metadata_semantic_tags (
    id INTEGER PRIMARY KEY,
    user_metadata_id INTEGER NOT NULL,
    tag_id INTEGER NOT NULL,
    applied_context TEXT,
    applied_variant TEXT,
    confidence REAL DEFAULT 1.0,
    source TEXT DEFAULT 'user',
    instance_attributes JSON,
    created_at TIMESTAMP NOT NULL,
    updated_at TIMESTAMP NOT NULL,
    device_uuid UUID NOT NULL,

    FOREIGN KEY (user_metadata_id) REFERENCES user_metadata(id) ON DELETE CASCADE,
    FOREIGN KEY (tag_id) REFERENCES semantic_tags(id) ON DELETE CASCADE,
    UNIQUE(user_metadata_id, tag_id)
);

-- Usage pattern tracking for intelligent suggestions
CREATE TABLE tag_usage_patterns (
    id INTEGER PRIMARY KEY,
    tag_id INTEGER NOT NULL,
    co_occurrence_tag_id INTEGER NOT NULL,
    occurrence_count INTEGER DEFAULT 1,
    last_used_together TIMESTAMP NOT NULL,

    FOREIGN KEY (tag_id) REFERENCES semantic_tags(id) ON DELETE CASCADE,
    FOREIGN KEY (co_occurrence_tag_id) REFERENCES semantic_tags(id) ON DELETE CASCADE,
    UNIQUE(tag_id, co_occurrence_tag_id)
);

-- Full-text search support
CREATE VIRTUAL TABLE tag_search_fts USING fts5(
    tag_id,
    canonical_name,
    display_name,
    formal_name,
    abbreviation,
    aliases,
    description,
    namespace,
    content='semantic_tags',
    content_rowid='id'
);

Closure Table Pattern

The closure table enables O(1) hierarchical queries by pre-computing all ancestor-descendant relationships:

-- Example: Technology → Programming → Web Development → React
-- Direct relationships:
INSERT INTO tag_relationships VALUES (1, 2, 'parent_child', 1.0); -- Tech → Programming
INSERT INTO tag_relationships VALUES (2, 3, 'parent_child', 1.0); -- Programming → Web Dev
INSERT INTO tag_relationships VALUES (3, 4, 'parent_child', 1.0); -- Web Dev → React

-- Closure table automatically maintains all paths:
INSERT INTO tag_closure VALUES (1, 1, 0, 1.0); -- Tech → Tech (self)
INSERT INTO tag_closure VALUES (1, 2, 1, 1.0); -- Tech → Programming
INSERT INTO tag_closure VALUES (1, 3, 2, 1.0); -- Tech → Web Dev (via Programming)
INSERT INTO tag_closure VALUES (1, 4, 3, 1.0); -- Tech → React (via Programming, Web Dev)
-- ... and so on for all relationships

This enables efficient queries like “find all content tagged with any descendant of Technology”:

SELECT DISTINCT e.*
FROM entries e
JOIN user_metadata_semantic_tags umst ON e.metadata_id = umst.user_metadata_id
JOIN tag_closure tc ON umst.tag_id = tc.descendant_id
WHERE tc.ancestor_id = (SELECT id FROM semantic_tags WHERE canonical_name = 'Technology');

Key Features

1. Polymorphic Naming

Multiple tags can share the same canonical name when differentiated by namespace:

// Same name, different contexts
let phoenix_city = SemanticTag {
    canonical_name: "Phoenix".to_string(),
    namespace: Some("Geography".to_string()),
    description: Some("City in Arizona, USA".to_string()),
    // ...
};

let phoenix_myth = SemanticTag {
    canonical_name: "Phoenix".to_string(),
    namespace: Some("Mythology".to_string()),
    description: Some("Mythical bird that rises from ashes".to_string()),
    // ...
};

This allows natural, human-friendly naming without forcing artificial uniqueness.

2. Semantic Variants

Each tag supports multiple access points for flexible user interaction:

let js_tag = SemanticTag {
    canonical_name: "JavaScript".to_string(),
    formal_name: Some("JavaScript Programming Language".to_string()),
    abbreviation: Some("JS".to_string()),
    aliases: vec!["ECMAScript".to_string(), "ES".to_string()],
    namespace: Some("Technology".to_string()),
    // ...
};

// All of these resolve to the same tag:
assert!(js_tag.matches_name("JavaScript"));
assert!(js_tag.matches_name("js"));          // Case insensitive
assert!(js_tag.matches_name("ECMAScript"));
assert!(js_tag.matches_name("JavaScript Programming Language"));

3. Context-Aware Resolution

When users type ambiguous tag names, the system intelligently resolves them based on existing context:

// User is working with geographic data and types "Phoenix"
let context_tags = vec![arizona_tag, usa_tag, city_tag];
let resolved = tag_resolver.resolve_ambiguous_tag("Phoenix", &context_tags).await?;
// Returns "Geography::Phoenix" (city) rather than "Mythology::Phoenix" (bird)

The resolution considers:

Namespace compatibility with existing tags
Usage patterns from historical co-occurrence
Hierarchical relationships between tags

4. Hierarchical Organization

Tags form a directed acyclic graph (DAG) structure supporting:

Technology
├── Programming
│   ├── Web Development
│   │   ├── Frontend
│   │   │   ├── React
│   │   │   └── Vue
│   │   └── Backend
│   │       ├── Node.js
│   │       └── Python
│   └── Mobile Development
│       ├── iOS
│       └── Android
└── Design
    ├── UI/UX
    └── Graphic Design

Benefits of hierarchical organization:

Implicit Classification: Tagging with “React” automatically inherits “Frontend”, “Web Development”, etc.
Semantic Discovery: Searching “Technology” surfaces all descendant content
Emergent Patterns: System reveals organizational connections users didn’t explicitly create

5. AI Integration

The system supports AI-powered tagging with confidence scoring:

// AI analyzes image and applies tags
let ai_application = TagApplication {
    tag_id: vacation_tag_id,
    applied_context: Some("image_analysis".to_string()),
    confidence: 0.92,
    source: TagSource::AI,
    instance_attributes: hashmap! {
        "detected_objects".to_string() => json!(["dog", "beach", "sunset"]),
        "model_version".to_string() => json!("v2.1")
    },
    // ...
};

AI features:

Confidence Scoring: 0.0-1.0 confidence levels for AI suggestions
User Review: Low confidence tags require user approval
Learning Loop: User corrections improve future AI suggestions
Privacy Options: Local models (Ollama) or cloud APIs with user control

6. Union Merge Conflict Resolution

During synchronization, tag conflicts are resolved using an additive approach:

// Device A: Photo tagged with "vacation"
let local_apps = vec![TagApplication::user_applied(vacation_tag_id, device_a)];

// Device B: Same photo tagged with "family"
let remote_apps = vec![TagApplication::user_applied(family_tag_id, device_b)];

// Union merge result: Photo tagged with BOTH "vacation" AND "family"
let merged = resolver.merge_tag_applications(local_apps, remote_apps).await?;

This prevents data loss and preserves all user intent during synchronization.

Manager Layer

TagManager

Core manager providing high-level tag operations. Located in ops/tags/manager.rs:

use crate::ops::tags::manager::TagManager;

impl TagManager {
    // Create new semantic tag
    pub async fn create_tag(
        &self,
        canonical_name: String,
        namespace: Option<String>,
        created_by_device: Uuid,
    ) -> Result<SemanticTag, TagError>;

    // Find tags by name (including variants)
    pub async fn find_tags_by_name(&self, name: &str) -> Result<Vec<SemanticTag>, TagError>;

    // Resolve ambiguous tag names using context
    pub async fn resolve_ambiguous_tag(
        &self,
        tag_name: &str,
        context_tags: &[SemanticTag],
    ) -> Result<Vec<SemanticTag>, TagError>;

    // Create hierarchical relationship
    pub async fn create_relationship(
        &self,
        parent_id: Uuid,
        child_id: Uuid,
        relationship_type: RelationshipType,
        strength: Option<f32&gt;,
    ) -> Result<(), TagError>;

    // Get all descendant tags
    pub async fn get_descendants(&self, tag_id: Uuid) -> Result<Vec<SemanticTag>, TagError>;

    // Discover organizational patterns
    pub async fn discover_organizational_patterns(&self) -> Result<Vec<OrganizationalPattern>, TagError>;

    // Merge tag applications (for sync)
    pub async fn merge_tag_applications(
        &self,
        local: Vec<TagApplication>,
        remote: Vec<TagApplication>,
    ) -> Result<TagMergeResult, TagError>;
}

TagContextResolver

Handles intelligent disambiguation of ambiguous tag names:

impl TagContextResolver {
    pub async fn resolve_ambiguous_tag(
        &self,
        tag_name: &str,
        context_tags: &[SemanticTag],
    ) -> Result<Vec<SemanticTag>, TagError> {
        let candidates = self.find_all_name_matches(tag_name).await?;

        if candidates.len() <= 1 {
            return Ok(candidates);
        }

        // Score candidates based on context compatibility
        let mut scored_candidates = Vec::new();
        for candidate in candidates {
            let mut score = 0.0;

            // Namespace compatibility
            score += self.calculate_namespace_compatibility(&candidate, context_tags).await?;

            // Usage pattern compatibility
            score += self.calculate_usage_compatibility(&candidate, context_tags).await?;

            // Hierarchical relationship compatibility
            score += self.calculate_hierarchy_compatibility(&candidate, context_tags).await?;

            scored_candidates.push((candidate, score));
        }

        // Return candidates sorted by relevance score
        scored_candidates.sort_by(|a, b| b.1.partial_cmp(&a.1).unwrap());
        Ok(scored_candidates.into_iter().map(|(tag, _)| tag).collect())
    }
}

TagUsageAnalyzer

Tracks usage patterns and discovers emergent organizational structures:

impl TagUsageAnalyzer {
    // Record when tags are used together
    pub async fn record_usage_patterns(
        &self,
        tag_applications: &[TagApplication],
    ) -> Result<(), TagError>;

    // Find frequently co-occurring tag pairs
    pub async fn get_frequent_co_occurrences(
        &self,
        min_count: i32,
    ) -> Result<Vec<(Uuid, Uuid, i32)>, TagError>;

    // Calculate how often a tag appears with context tags
    pub async fn calculate_co_occurrence_score(
        &self,
        candidate: &SemanticTag,
        context_tags: &[SemanticTag],
    ) -> Result<f32, TagError>;
}

UserMetadataManager

Manages user metadata including semantic tag applications. Located in ops/metadata/manager.rs:

use crate::ops::metadata::manager::UserMetadataManager;

impl UserMetadataManager {
    // Apply semantic tags to user metadata
    pub async fn apply_semantic_tags(
        &self,
        entry_uuid: Uuid,
        tag_applications: Vec<TagApplication>,
        device_id: Uuid,
    ) -> Result<(), TagError>;

    // Get all tags applied to an entry
    pub async fn get_applied_tags(
        &self,
        entry_uuid: Uuid,
    ) -> Result<Vec<TagApplication>, TagError>;

    // Remove tags from an entry
    pub async fn remove_tags(
        &self,
        entry_uuid: Uuid,
        tag_ids: Vec<Uuid>,
    ) -> Result<(), TagError>;
}

Usage Examples

Basic Tag Creation

use crate::ops::tags::manager::TagManager;
use std::sync::Arc;

let manager = TagManager::new(Arc::new(db.conn().clone()));

// Create a basic tag
let project_tag = manager.create_tag(
    "Project".to_string(),
    None,
    device_id
).await?;

// Create contextual tags
let phoenix_city = manager.create_tag(
    "Phoenix".to_string(),
    Some("Geography".to_string()),
    device_id
).await?;

let phoenix_myth = manager.create_tag(
    "Phoenix".to_string(),
    Some("Mythology".to_string()),
    device_id
).await?;

Building Hierarchies

// Create tag hierarchy: Technology → Programming → Web Development
let tech_tag = manager.create_tag("Technology".to_string(), None, device_id).await?;
let prog_tag = manager.create_tag("Programming".to_string(), None, device_id).await?;
let web_tag = manager.create_tag("Web Development".to_string(), None, device_id).await?;

// Create parent-child relationships
manager.create_relationship(
    tech_tag.id,
    prog_tag.id,
    RelationshipType::ParentChild,
    None
).await?;

manager.create_relationship(
    prog_tag.id,
    web_tag.id,
    RelationshipType::ParentChild,
    None
).await?;

// Query descendants
let all_tech_tags = manager.get_descendants(tech_tag.id).await?;
// Returns: [Programming, Web Development, and any other descendant tags]

Applying Tags to Content

// User manually tags a file
let user_app = TagApplication::user_applied(javascript_tag_id, device_id);

// AI analyzes and suggests tags
let ai_app = TagApplication::ai_applied(react_tag_id, 0.95, device_id);
ai_app.applied_context = Some("code_analysis".to_string());

// Apply tags to user metadata
let applications = vec![user_app, ai_app];
manager.record_tag_usage(&applications).await?;

Context Resolution

// User types "JS" while working with React files
let context_tags = vec![react_tag, frontend_tag, web_dev_tag];
let resolved = manager.resolve_ambiguous_tag("JS", &context_tags).await?;
// Returns JavaScript tag (in Technology namespace) as best match

Pattern Discovery

// Discover emergent organizational patterns
let patterns = manager.discover_organizational_patterns().await?;

for pattern in patterns {
    match pattern.pattern_type {
        PatternType::FrequentCoOccurrence => {
            println!("Tags often used together: suggest relationship");
        }
        PatternType::HierarchicalRelationship => {
            println!("Suggest parent-child relationship");
        }
        PatternType::ContextualGrouping => {
            println!("Suggest namespace grouping");
        }
    }
}

Integration with Core Systems

Entry-Centric Metadata

Every Entry has immediate metadata capability through the metadata_id field:

// Entry always links to UserMetadata
pub struct Entry {
    pub metadata_id: i32,  // Always present - immediate tagging!
    // ... other fields
}

// UserMetadata contains semantic tag applications
pub struct UserMetadata {
    pub semantic_tags: Vec<TagApplication>,  // Enhanced tag applications
    // ... other metadata
}

Action System Integration

The semantic tagging system integrates with Spacedrive’s Action System for validation, audit logging, and transactional operations:

// Tag creation through actions
use crate::ops::tags::create::{CreateTagAction, CreateTagInput};

let action = CreateTagAction::new(CreateTagInput {
    canonical_name: "JavaScript".to_string(),
    namespace: Some("Technology".to_string()),
    // ... other fields
});

let result = action.execute(library, context).await?;

// Tag application through actions
use crate::ops::tags::apply::{ApplyTagsAction, ApplyTagsInput};

let action = ApplyTagsAction::new(ApplyTagsInput {
    entry_ids: vec![entry_id],
    tag_applications: vec![tag_application],
});

let result = action.execute(library, context).await?;

This enables:

Instant Tagging: Files can be tagged immediately upon discovery
Rich Context: Each tag application includes confidence, source, and attributes
Sync Integration: Tag applications sync with conflict resolution

Indexing System Integration

The indexing system can trigger automatic tagging during the Intelligence Queueing Phase:

// During indexing, queue AI analysis jobs
if entry.kind == EntryKind::File {
    match entry.file_type {
        FileType::Image => {
            job_queue.push(ImageAnalysisJob::new(entry.id)).await?;
        }
        FileType::Code => {
            job_queue.push(CodeAnalysisJob::new(entry.id)).await?;
        }
        // ... other types
    }
}

AI analysis jobs apply semantic tags with confidence scores.

Search Integration

The Temporal-Semantic Search system leverages semantic tags for enhanced discovery:

-- Semantic search using tag hierarchy
SELECT DISTINCT e.*
FROM entries e
JOIN user_metadata_semantic_tags umst ON e.metadata_id = umst.user_metadata_id
JOIN tag_closure tc ON umst.tag_id = tc.descendant_id
JOIN semantic_tags st ON tc.ancestor_id = st.id
WHERE st.canonical_name = 'Technology'
  AND umst.confidence > 0.8;

This enables queries like “find all Technology-related content” to surface files tagged with any descendant technology tags.

Sync System Integration

Semantic tags integrate with Library Sync using union merge resolution:

// Tags sync in the UserMetadata domain
impl Syncable for UserMetadataSemanticTag {
    fn get_sync_domain(&self) -> SyncDomain {
        SyncDomain::UserMetadata  // Union merge strategy
    }
}

// Conflict resolution preserves all tags
let merged_tags = resolver.merge_tag_applications(
    local_applications,
    remote_applications
).await?;

Performance Considerations

Closure Table Benefits

The closure table pattern provides O(1) hierarchical queries:

Ancestor Queries: SELECT * FROM tag_closure WHERE descendant_id = ?
Descendant Queries: SELECT * FROM tag_closure WHERE ancestor_id = ?
Path Queries: SELECT * FROM tag_closure WHERE ancestor_id = ? AND descendant_id = ?
Depth Queries: SELECT * FROM tag_closure WHERE depth = ?

Indexing Strategy

Key database indexes for performance:

-- Tag lookup indexes
CREATE INDEX idx_semantic_tags_canonical_name ON semantic_tags(canonical_name);
CREATE INDEX idx_semantic_tags_namespace ON semantic_tags(namespace);
CREATE INDEX idx_semantic_tags_type ON semantic_tags(tag_type);
CREATE INDEX idx_semantic_tags_privacy ON semantic_tags(privacy_level);

-- Closure table indexes
CREATE INDEX idx_tag_closure_ancestor ON tag_closure(ancestor_id);
CREATE INDEX idx_tag_closure_descendant ON tag_closure(descendant_id);
CREATE INDEX idx_tag_closure_depth ON tag_closure(depth);

-- Application indexes
CREATE INDEX idx_user_metadata_semantic_tags_metadata ON user_metadata_semantic_tags(user_metadata_id);
CREATE INDEX idx_user_metadata_semantic_tags_tag ON user_metadata_semantic_tags(tag_id);
CREATE INDEX idx_user_metadata_semantic_tags_source ON user_metadata_semantic_tags(source);

Full-Text Search

SQLite FTS5 provides efficient text search across all tag variants:

-- Search across all tag text fields
SELECT tag_id, rank FROM tag_search_fts
WHERE tag_search_fts MATCH 'javascript OR js OR ecmascript'
ORDER BY rank;

File Organization

The semantic tagging system is organized in the ops/ directory following Spacedrive’s architectural patterns:

core/src/ops/
├── tags/
│   ├── manager.rs                   # Core tag management logic
│   ├── facade.rs                    # High-level facade for UI/CLI
│   ├── apply/                       # Tag application actions
│   │   └── action.rs
│   ├── create/                      # Tag creation actions
│   │   └── action.rs
│   └── search/                      # Tag search actions
│       └── action.rs
└── metadata/
    └── manager.rs     # User metadata management

Migration Strategy

Since this is a development codebase with no existing users, the semantic tagging system completely replaces the old simple tag system:

Database Migration: m20250115_000001_semantic_tags.rs creates all new tables
Clean Implementation: No data migration or backward compatibility needed
Feature Complete: All whitepaper features available from day one
Performance Optimized: Built with proper indexing and closure table
Action Integration: Full integration with Spacedrive’s Action System

Future Enhancements

Planned advanced features building on this foundation:

Enterprise RBAC Integration

// Role-based access control for tags
pub struct TagPermission {
    pub role: UserRole,
    pub tag_namespace: Option<String>,
    pub operations: Vec<TagOperation>, // Create, Read, Update, Delete, Apply
}

Advanced AI Features

Semantic Similarity: Vector embeddings for content-based tag suggestions
Temporal Patterns: Time-based usage analysis for lifecycle tagging
Cross-Library Learning: Federated learning across user libraries (privacy-preserving)

Enhanced Sync Features

Selective Sync: Choose which tag namespaces to sync across devices
Conflict Policies: User-configurable resolution strategies
Audit Trail: Complete history of tag operations across all devices

This semantic tagging architecture transforms Spacedrive from having simple labels to providing a sophisticated knowledge management foundation that scales from personal use to enterprise deployment.

​Overview

​Core Architecture

​Design Principles

​Core Components

​Data Models

​SemanticTag

​TagType Enum

​PrivacyLevel Enum

​TagRelationship

​TagApplication

​Database Schema

​Tables Overview

​Closure Table Pattern

​Key Features

​1. Polymorphic Naming

​2. Semantic Variants

​3. Context-Aware Resolution

​4. Hierarchical Organization

​5. AI Integration

​6. Union Merge Conflict Resolution

​Manager Layer

​TagManager

​TagContextResolver

​TagUsageAnalyzer

​UserMetadataManager

​Usage Examples

​Basic Tag Creation

​Building Hierarchies

​Applying Tags to Content

​Context Resolution

​Pattern Discovery

​Integration with Core Systems

​Entry-Centric Metadata

​Action System Integration

​Indexing System Integration

​Search Integration

​Sync System Integration

​Performance Considerations

​Closure Table Benefits

​Indexing Strategy

​Full-Text Search

​File Organization

​Migration Strategy

​Future Enhancements

​Enterprise RBAC Integration

​Advanced AI Features

​Enhanced Sync Features