Hello, I’m currently experimenting with a new approach to code completion in Pharo. Unlike the existing token-based completion (which suggests one token at a time), this approach uses an LLM to generate larger, meaningful code segments—also known as Fill-in-the-Middle (FIM) code completion. This message is simply to gauge interest in this topic.
Once the system is ready, we will reach out to interested users. There is nothing specific expected from you—just using the plugin as you normally code. The plugin will handle the evaluation process and, with your consent, send us a report summarizing usage. This report will only include aggregated metrics (such as acceptance and rejection rates). We will not collect any information about your code (e.g., class names, method names, or content). The plugin itself is lightweight and runs entirely on your machine. The LLM is also designed to run locally, with our largest model requiring approximately 5 GB of RAM. If you are interested in this new approach to code completion, feel free to reply to this email. Best regards, Omar AbedelKader
