Project-aware agentic coding, validation and rollback.
Overview
AevCode is two things: the scripting language inside the engine, and the agentic pipeline that plans, validates and rolls back changes to a project. The second is the reason the first matters, an agent that can write code into your project is only safe if every change it makes is reviewable and reversible.
Where it fits
AevCode is one system in a pipeline. Each stage takes a defined input and hands a defined output to the next.
Four systems in one production path: Control, Orchestrates the project, the workflows and the agents. AevCode, Plans changes, validates them, and rolls them back cleanly when they fail. Design, Produces the images, textures, materials, meshes and animation a change needs. Core, Renders, simulates, edits and runs the project.
Orchestrates the project, the workflows and the agents.
Control decides what needs to happen and calls AevCode to work out how.
Plans changes, validates them, and rolls them back cleanly when they fail.
Validated changes are applied to the project Core builds and runs.
Produces the images, textures, materials, meshes and animation a change needs.
Assets land in the project alongside the change that asked for them.
Renders, simulates, edits and runs the project.
AevCode is the engine scripting language, verified as such in the built editor rather than presented as a rename. Visual graphs are a live interpreter over the same runtime, with breakpoints, stepping and per-node profiling, not a code generator that produces script you then maintain by hand.
Evidence
Each item names its source and the date it was last checked.
AevCode is the scripting layer inside the built editor, not a wrapper around one. Scripts report AevCode as their class in the running engine.
Source: Headless run against the built editor, reading the script class back from the engine · Verified
Related
