How to Run Inference in Codex with the Roboflow Plugin

Roboflow’s plugin for OpenAI Codex enables users to find trained models in their Roboflow workspace or on Roboflow Universe, run object detection or instance segmentation on images through natural-language prompts, and generate visualizations and analytics. The tutorial explains plugin installation and account connection, including browser-based authentication and a project-listing test, then demonstrates three workflows: ceramic tile defect detection using a workspace model, steel surface-defect detection using a public RF-DETR model, and pallet instance segmentation using an RF-DETR segmentation model. Each workflow covers selecting and verifying a model, attaching an original image, running inference with a confidence threshold, saving full prediction data, and creating annotated images or reports with consistent labels, charts, tables, and downloadable files. It emphasizes reviewing predictions for errors, treating confidence and image-level counts appropriately, retaining original class and geometry data, and evaluating models on representative labeled images before operational use. After testing, users can deploy models through Roboflow’s hosted API, configurable Workflows, or self-hosted inference on local and edge hardware.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.