Blender forbids Python threads. So my AI add-on talks to itself through two JSON files
The pivot I spent the last months building a local multi-agent CLI code generator - planner → coder → tester → executor → reviewer, running on open models via Ollama. It closed the loop: it wrote code, ran it, tested it, and fixed it. It worked. And that's exactly the problem: it solved a problem I don't have. Writing code for files on disk is a solved niche - I was competing with every other…
The AI add-on called Game Asset Assistant, which runs inside Blender 4.2+, allows users to interact with 3D models using an AI assistant. The assistant accesses the current scene, objects, and their properties, and sends a concise snapshot to the AI model via JSON files. The model responds with code blocks that can be executed in the Blender main thread, allowing for a seamless loop of code generation, execution, and error resolution.
The add-on achieves this by communicating with the AI model through JSON files, instead of using Python threads, which Blender prohibits from running outside the main thread due to potential crashes in the drawing code. The development of the add-on is divided into several phases, with the first three focusing on richer scene snapshots, undo functionality, and destructive operation confirmations.
The roadmap also includes features like texturing, UV manipulation, auto-rigging, and one-click pipelines. The developer plans to release progress updates and invites collaboration from others working on AI integration in 3D pipelines.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
