Module 09 · Frameworks

State graphs and durable execution

Rebuild what a graph runtime gives you: channels with reducers, a super-step scheduler that knows which nodes are stale, and a checkpointer that makes resume, time travel and forking fall out of one dict.

4 lessons · 3 graded exercises·4.5 hours·intermediate·Prerequisite: Agent frameworks

Why a graph beats a loop

25 min

Where the graph runtime sits underneath higher level harnesses, what branching and durability buy, and when a graph adds unnecessary ceremony around a while loop.

Channels and reducers

45 min · exercise

Graph state relies on named channels with dedicated reducers and version counters that determine node execution.

The super step

60 min · exercise

Nodes eligible in the same step read the same snapshot before their updates merge. This rule makes fan-out and fan-in deterministic.

Checkpoints, resume, and time travel

55 min · exercise

Persisting state after every super-step enables crash recovery, historical replay, and thread branching from saved state records.