Getting Started

In-Memory Store

The simplest way to use SemanticSpacetime.jl requires no database at all. MemoryStore is ideal for experimentation, testing, and lightweight applications.

using SemanticSpacetime

# Register built-in arrow types
add_mandatory_arrows!()

# Create an in-memory store
store = MemoryStore()

# Create nodes
n1 = mem_vertex!(store, "hello", "greetings")
n2 = mem_vertex!(store, "world", "greetings")

# Create edges (arrows must be registered)
mem_edge!(store, n1, "then", n2)

# Query
println(node_count(store))   # 2
println(link_count(store))   # 2 (forward + inverse)

# Search
results = mem_search_text(store, "hello")

Database-Backed Store

For persistent storage, load a database extension. The DBStore uses DBInterface.jl so the same API works across backends.

SQLite

using SemanticSpacetime, SQLite

# In-memory SQLite store
store = open_sqlite()

# File-backed SQLite store
store = open_sqlite("knowledge.db")

DuckDB

using SemanticSpacetime, DuckDB

store = open_duckdb()               # in-memory
store = open_duckdb("knowledge.db") # file-backed

PostgreSQL

For large-scale deployments, the original PostgreSQL backend is available:

using SemanticSpacetime

sst = open_sst()  # connects via CREDENTIALS_FILE or defaults

# Create nodes and links
n1 = vertex!(sst, "Mary had a little lamb", "nursery rhymes")
n2 = vertex!(sst, "Its fleece was white as snow", "nursery rhymes")
edge!(sst, n1, "then", n2, String[], 1.0f0)

close_sst(sst)

Arrow Types

Arrows are named, typed relationships. They must be registered before use. The SST system provides mandatory arrows via add_mandatory_arrows!, and you can register custom ones:

# Register a forward and inverse arrow pair
fwd = insert_arrow!("leadsto", "causes", "causes outcome", "+")
bwd = insert_arrow!("leadsto", "caused-by", "is caused by", "-")
insert_inverse_arrow!(fwd, bwd)

Contexts

Contexts scope nodes and links. They are sorted, comma-joined string labels:

ctx_ptr = register_context!(["biology", "genetics"])
ctx_str = get_context(ctx_ptr)  # "biology,genetics"

Text Size Classes

Nodes are bucketed by text length for efficient lookup:

Use n_channel to determine the class for a given string.