What are Triples?
Triples are the building blocks of your knowledge graph. Each triple represents a fact in the form:
Subject → Predicate → Object
For example:
“Tesla” → “headquarters_in” → “Austin, Texas”
“Elon Musk” → “founded” → “SpaceX”
“SpaceX” → “founded_year” → “2002”
Creating Triples
Single Triple
curl -X POST https://neode.ai/api/triples \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"subject": "Tesla",
"predicate": "headquarters_in",
"object": "Austin, Texas",
"object_type": "entity",
"confidence": 0.95,
"source": "https://wikipedia.org/wiki/Tesla",
"index_id": "YOUR_INDEX_ID"
}'
Batch Creation
Create multiple triples in one request:
curl -X POST https://neode.ai/api/triples \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"triples": [
{
"subject": "Tesla",
"predicate": "founded_year",
"object": "2003",
"object_type": "literal"
},
{
"subject": "Tesla",
"predicate": "industry",
"object": "Electric Vehicles",
"object_type": "entity"
},
{
"subject": "Tesla",
"predicate": "ceo",
"object": "Elon Musk",
"object_type": "entity"
}
],
"index_id": "YOUR_INDEX_ID",
"confidence": 0.9
}'
Batch creation is more efficient than individual requests. Use it when storing multiple facts at once.
Querying Triples
By Subject
Get all facts about a specific entity:
curl "https://neode.ai/api/triples?subject=Tesla" \
-H "Authorization: Bearer YOUR_API_KEY"
By Predicate
Get all facts of a specific type:
curl "https://neode.ai/api/triples?predicate=founded_by" \
-H "Authorization: Bearer YOUR_API_KEY"
By Entity ID
Query using the resolved entity IDs:
# Get triples where entity is the subject
curl "https://neode.ai/api/triples?subject_entity_id=ENTITY_UUID" \
-H "Authorization: Bearer YOUR_API_KEY"
# Get triples where entity is the object
curl "https://neode.ai/api/triples?object_entity_id=ENTITY_UUID" \
-H "Authorization: Bearer YOUR_API_KEY"
Full-Text Search
Search across subject, predicate, and object:
curl "https://neode.ai/api/triples?search=electric%20vehicle" \
-H "Authorization: Bearer YOUR_API_KEY"
Filter by Graph
Get triples from a specific graph:
curl "https://neode.ai/api/triples?graph_id=GRAPH_UUID" \
-H "Authorization: Bearer YOUR_API_KEY"
Combining Filters
curl "https://neode.ai/api/triples?subject=Tesla&predicate=founded_by&limit=10" \
-H "Authorization: Bearer YOUR_API_KEY"
Updating Triples
Update an existing triple by ID:
curl -X PUT https://neode.ai/api/triples \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"id": "TRIPLE_UUID",
"confidence": 0.99,
"source": "https://updated-source.com"
}'
You can update:
subject, predicate, object
confidence
source
graph_id
Deleting Triples
Single Delete
curl -X DELETE "https://neode.ai/api/triples?id=TRIPLE_UUID" \
-H "Authorization: Bearer YOUR_API_KEY"
Batch Delete
curl -X DELETE "https://neode.ai/api/triples?ids=UUID1,UUID2,UUID3" \
-H "Authorization: Bearer YOUR_API_KEY"
Best Practices
Predicate Naming
Use consistent, lowercase, snake_case predicates:
Good Avoid founded_byFoundedBy, Founded Byheadquarters_inheadquartersIn, HQbirth_datebirthDate, DOB
Confidence Scores
Be realistic with confidence scores:
{
"confidence" : 0.95 // Verified from authoritative source
}
{
"confidence" : 0.7 // From news article, may change
}
Sources
Always include sources when possible:
{
"source" : "https://wikipedia.org/wiki/Tesla" ,
"confidence" : 0.95
}
Entity vs Literal
Use the correct object type:
// Entity - references another thing
{
"object" : "Elon Musk" ,
"object_type" : "entity"
}
// Literal - a value
{
"object" : "2003" ,
"object_type" : "literal"
}
Response Example
{
"success" : true ,
"data" : {
"id" : "550e8400-e29b-41d4-a716-446655440000" ,
"subject" : "Tesla" ,
"predicate" : "founded_by" ,
"object" : "Elon Musk" ,
"object_type" : "entity" ,
"confidence" : 0.95 ,
"source" : "https://wikipedia.org/wiki/Tesla" ,
"subject_entity_id" : "entity-uuid-1" ,
"object_entity_id" : "entity-uuid-2" ,
"index_id" : "index-uuid" ,
"created_at" : "2026-01-27T10:00:00Z"
}
}
API Reference See the complete Triples API documentation.