A time-aware directed multigraph for evolving dialogue, belief, and memory structures.
The script real_run3.py logs tutoring turns and exports paper-oriented metrics, plus external baselines (LLM-only tutor script; BKT, DKT-style, and heuristic KT controllers) and ablations (no policy, no temporal updates, single-node graph). See How to run experiments, Simulation metrics (real_run3), Formal mechanics (LaTeX / reproducibility), and Paper notes: KT positioning.
from dialograph import Dialograph, Node, Edge
g = Dialograph()
a = Node(node_id="a", node_type="intent")
b = Node(node_id="b", node_type="belief")
g.add_node(a)
g.add_node(b)
g.add_edge(
Edge(
source_node_id="a",
target_node_id="b",
relation="elicits"
)
)