JetBrains has rolled out version 2024.1 for all IDEs under its umbrella, introducing a new feature called full line code completion powered by AI that predicts the next line of code we are likely to type in the editor, with the ability to work offline without needing an internet connection.
The functionality of full line code completion displays grayed-out text in the following line and allows users to press tab to have the editor complete that line for them. This feature supports code in languages such as Java, Kotlin, Python, JavaScript, TypeScript, CSS, PHP, Go, and Ruby.
Operating locally without requiring an internet connection, this feature does not send data outside of the machine. Its backend operation is based on language models trained specifically for each programming language separately, using data from open-source code. The model runs in a separate process from the IDE itself.
JetBrains states that opting to insert code line by line is aimed at enabling users to review code faster than inserting lengthy blocks of code all at once. On the other hand, the model is not overly large (with 100 million parameters, an input length of 1,536 tokens, or approximately 170 lines of code), striking a suitable balance between efficiency and usability.
This feature differs from JetBrains AI Assistant, which can perform multi-line code completion but requires cloud-based operation and additional package purchase beyond the regular IDE license cost, whereas full line code completion is included in the IDE’s price.
TLDR: JetBrains introduces full line code completion feature in version 2024.1 for all IDEs, utilizing AI to predict and complete the next line of code offline, supporting various programming languages while operating locally and training on open-source code.
Leave a Comment