Refact.ai: GPU server setup and IDE connection

This tutorial, uploaded on May 30, 2024, explains how to set up Refact.ai as a self-hosted coding assistant using your own GPU. The source gives no recording date or Refact.ai version. Its Docker workflow and memory recommendation describe the setup at the time of recording.

The speaker breaks down a Docker command rather than explaining each installation prerequisite. The GPU flag selects the devices available to the container, while the shared-memory flag controls its allocation. The speaker recommends a value of 256 at recording time and warns that more memory may be needed later; the transcript does not specify the unit. A volume flag provides persistent storage for settings, downloaded models and fine-tuning results.

The update procedure stops the running container, pulls the Docker image and restarts the container. To connect the IDE plugin, the tutorial selects the self-hosted login option and enters a localhost URL with the port chosen during setup. It does not name a specific IDE.

The speaker describes Refact.ai as open source. The video focuses on server setup and plugin connection; it does not test completion quality or report GPU performance.