GitHub

Search and Path Solving

SemanticSpacetime.jl provides several search and traversal strategies.

Text Search

In-Memory

results = mem_search_text(store, "lamb")

Database

ptrs = search_nodes(sst, "lamb", "nursery rhymes")
results = search_text(sst, "lamb")

The database search uses PostgreSQL tsvector full-text indexing for fast lookups.

A causal cone is the set of all nodes reachable by following links forward (future cone) or backward (past cone) from a starting node.

# Forward cone — what follows from this node?
cone = forward_cone(store, start_nptr; max_depth=5)

# Backward cone — what leads to this node?
cone = backward_cone(store, start_nptr; max_depth=5)

The result is a ConeResult containing paths and supernodes (nodes appearing in multiple paths).

Path Solving

Find paths between two nodes:

result = find_paths(store, from_nptr, to_nptr; max_depth=10)
println(result.dag_paths)     # acyclic paths
println(result.loop_paths)    # paths containing cycles

Detect loops within paths:

loops = detect_path_loops(paths)

BFS with weight filtering and Dijkstra's shortest path:

# BFS with minimum weight threshold
paths = weighted_search(store, start_nptr; max_depth=5, min_weight=0.5)

# Shortest weighted path (weight = inverse distance)
path = dijkstra_path(store, from_nptr, to_nptr)

# Rank paths by total weight
ranked = rank_by_weight(paths; descending=true)

Story Selection

Follow sequences (stories) through the graph by arrow type:

stories = select_stories_by_arrow(store, start_nptr, arrow_ptr)

Inhibition Context

Filter search results with include/exclude rules:

ctx = parse_inhibition_context("+biology,-chemistry")
filtered = search_with_inhibition(store, query, ctx)

A single entry point that dispatches to the appropriate search strategy:

params = UnifiedSearchParams(query="lamb", chapter="nursery rhymes")
results = unified_search(store, params)