Signals

Signals are the atomic unit of data in Kremis: Entity | Attribute | Value.

A signal is a grounded observation in the form:

Entity | Attribute | Value

Every piece of data in Kremis enters as a signal. There is no other way to add information to the graph.

Structure

Field	Type	Constraints	Description
entity_id	u64	—	Unique identifier for the entity
attribute	string	Max 256 bytes, non-empty	The attribute name
value	string	Max 64 KB, non-empty	The attribute value

Example

{"entity_id": 1, "attribute": "name", "value": "Alice"}

This signal records that entity 1 has an attribute name with value Alice.

Ingestion Behavior

When a signal is ingested:

1. Validate — attribute and value length are checked
2. Node creation — a node is created for the entity (or the existing one is reused)
3. Property storage — the attribute/value pair is stored as a property on the node
4. Edge formation — when signals are ingested in sequence, adjacent signals form edges with weight +1

Repeated signals on the same edge cause the weight to increment (saturating arithmetic). This is how patterns emerge — frequently co-occurring signals produce stronger edges.

Input Formats

JSON

[
  {"entity_id": 1, "attribute": "name", "value": "Alice"},
  {"entity_id": 2, "attribute": "name", "value": "Bob"},
  {"entity_id": 1, "attribute": "knows", "value": "Bob"}
]

Text

Colon-separated entity_id:attribute:value per line:

1:name:Alice
2:name:Bob
1:knows:Bob