Semantic Search¶
Archivus semantic search finds documents by meaning, not just keywords. Powered by the Knowledge Graph and vector embeddings, semantic search understands context, relationships, and intent to deliver precisely what you need.
How It Works¶
Traditional Keyword Search:
Search: "invoice"
Results: Documents containing the exact word "invoice"
Misses: Bills, statements, payment requests
Archivus Semantic Search:
Search: "invoice"
Results: Invoices, bills, statements, payment requests, receipts
Finds: Documents with similar MEANING
Search Technology¶
Starter+
Vector Embeddings
- Documents converted to 1536-dimension vectors
- Mathematical similarity matching
- OpenAI text-embedding-3-small model
- IVFFlat indexing for sub-second performance
Knowledge Graph Integration Enterprise
- Entity-aware search queries
- Relationship traversal in search results
- Temporal context filtering
- Provenance-based ranking
Search Types¶
Text Search¶
Free+
Fast keyword-based search across document content:
- Sub-100ms average response time
- Boolean operators (AND, OR, NOT)
- Wildcard support
- Filter by document type, tags, dates
Semantic Search¶
Starter+
Meaning-based search using AI embeddings:
- Sub-800ms average response time
- Finds conceptually similar documents
- Cross-language potential (future)
- 0.5 credits per query
Entity Search¶
Enterprise
Search by extracted entities and relationships:
- Find all documents mentioning an organization
- Discover documents with specific people or dates
- Filter by entity type and attributes
- Relationship-aware results
Example:
Search: "Acme Corporation"
Results: All documents containing:
- Acme Corp (exact)
- Acme Corporation (variation)
- Acme Co. (abbreviation)
- Related entities: Acme subsidiaries
Advanced Filters¶
Combine semantic search with powerful filters:
Document Attributes
- Document type (invoice, contract, report, etc.)
- Upload date range
- File size and format
- Processing status
Organization Hierarchy
- Workspace or project scope
- Collection membership
- Permission level
- Owner or contributor
AI Metadata
- Confidence scores
- Extracted entities
- Classification results
- AI-generated tags
Search Results¶
Rich Result Cards
Each result includes:
- Document preview with highlighted matches
- Extracted entities and key information
- Relevance score and ranking explanation
- Quick actions (view, download, analyze)
Result Ranking
Results ranked by:
- Semantic relevance to query
- Entity match quality (Enterprise)
- Recency (configurable weight)
- User permissions (only accessible docs)
Knowledge Graph Search¶
Enterprise
Search the Knowledge Graph directly:
Graph Queries
- Find entities and their relationships
- Traverse the graph (depth-limited)
- Discover indirect connections
- Query claim networks
Example Graph Query:
Query: "Who are Acme Corp's executives mentioned in our documents?"
Results:
- John Smith (CEO) - 12 documents
- Jane Doe (CFO) - 8 documents
- Bob Johnson (CTO) - 5 documents
Relationships:
- Reports to: Board of Directors
- Employed by: Acme Corporation
- Mentioned in: Q4 2025 Board Meeting Minutes
Search Performance¶
Scalability
- Handles 100,000+ documents
- Sub-second search across entire corpus
- Parallel query execution
- Efficient index maintenance
Optimization
- IVFFlat indexing for vector search
- PostgreSQL full-text search (text mode)
- Query result caching
- Automatic index updates
Search API¶
Programmatic search access via REST API:
# Semantic search
POST /api/v1/search/semantic
{
"query": "contract renewal deadlines",
"filters": {
"document_type": "CONTRACT",
"date_range": "2026-Q1"
},
"limit": 20
}
# Entity search (Enterprise)
POST /api/v1/search/entities
{
"entity_type": "ORGANIZATION",
"query": "Acme",
"include_relationships": true
}
Search Use Cases¶
Legal Research
- Find similar contracts or precedents
- Discover related case documents
- Search by party names or dates
Compliance
- Locate documents with specific entities
- Find contracts expiring in date range
- Search for regulatory keywords
Due Diligence
- Research company mentions
- Find financial documents
- Discover relationship networks
Knowledge Discovery
- Explore conceptually related content
- Find documents you didn't know existed
- Uncover hidden relationships
Search Analytics¶
Team+
Track search usage and effectiveness:
- Query patterns and frequency
- Result click-through rates
- Semantic vs. text search ratio
- Popular entities and topics
Integration¶
Search integrates with other capabilities:
- AI Assistant: Archie uses search to find context
- Rules Engine: Trigger actions based on search results
- Workflows: Search nodes in DAG pipelines
- Research: Ground findings against search results
Getting Started¶
Web Interface
- Navigate to the search bar
- Enter your query
- Apply filters as needed
- Review ranked results
API Access