Feiyu's Technical Blog
Notes on software, systems, and things I am learning.
Agents & Automation
Local models, tool-using agents, and practical information-extraction workflows.
How I Let Codex Talk to ChatGPT Through My Signed-In Chrome
A local browser bridge lets Codex send text or PDFs to a normal ChatGPT conversation, wait for the answer, and continue the same chat later.
Local LLM + Agent for Web Information Extraction
A local agent architecture for semantic web extraction, browser automation, bounded model context, and durable task state.
Game Perception — Olden Era
Computer-vision experiments that turn arena gameplay screenshots into structured game state.
Training a Game UI Icon Detector Using Only the Icon Catalog
We train on 390 catalog icons, then test whether the model can find and identify heroes and skills in complete game screenshots.
Why Five Game Icons Were Missed—and What Improved Them
We trace five missed detections through the full-screen icon pipeline, then compare calibration, NMS, type-specific outputs, and a spatial H2 head.
Full-Image Icon Search with Sliding Windows and Visual Embeddings
In this seven-frame experiment, the search often identifies the right hero or skill, but its highest-scoring box is not always the best crop.