Graph engineering
The design of stateful, branching execution graphs for agentic work.
Formal definition
Graph engineering. A canonical LLMography term. See the formal definition in the article below.
Formal definition
Graph engineering is the practice of structuring AI work as stateful graphs with branches, memory, and typed relations among actors.
In plain language
Graph engineering is not a dashboard widget. It is a way of naming a part of Human–AI work that otherwise disappears into logs, chat history, or an unexamined final answer.
Why it matters
As systems move from prompts to agents and loops, decisions are produced across many steps. Without this concept, those steps remain difficult to reconstruct, compare, or audit.
Diagram
The signature object of LLMography is a trajectory: human, model, tool, observation, verification, revision, decision. Graph engineering occupies a distinct position on that object.
Related metrics
See the proposed measures for loop autonomy, verification strength, and decision-trace completeness. These are research indicators, not certified scores.
Example
A recommendation is issued after a tool call fails verification and a human requests a narrower claim. Reconstructing that path is an instance of this concept in use.
Citation
Bousmah, M. LLMography concept note: Graph engineering. LLMography research site.