Artificial intelligence code completion tool provider Tabnine Ltd. today introduced a new more intuitive way for developers to complete AI-assisted coding tasks directly in the editor with inline actions that work directly on selected snippets of code.
The company’s flagship product is an AI-powered code completion tool for professional software developers in an integrated development environment, which helps save them time and energy by providing code suggestions as they type. The new tool will also join Tabnine Chat, which allows users to “talk” to their code using generative AI.
“Inline actions brings together the best of chat and code completions into a single interface,” the company said in the announcement. “By delivering AI-generated code or edits within the same environment where you’re working, inline actions are a faster and easier method to create, refine, document and fix code quickly.”
Developers need only select the code that needs to be worked with, pick a predefined action or ask Tabnine to complete the task. The results are completed inline – meaning that they happen within the code itself – with visual highlights of what has changed.
Developers can then accept, modify or reject the changes right from within the modified code snippet. The company said this allows developers to stay in the flow of coding using the editor without needing to make manual edits, which results in a smoother experience.
“By delivering AI-generated code or edits within the same environment where you’re working, inline actions are a faster and easier method to create, refine, document, and fix code quickly,” Tabnine said.
Inline actions are currently supported by Visual Studio Code and the JetBrains family of code editors, with support for other development environments coming soon.
Users of the new capability can flexibly choose between the company’s various switchable large language models. This allows customers to choose between different LLMs depending on their particular use case in enterprise environments, such as Tabnine’s custom-developed models, or from popular models from third parties such as Cohere Inc.’s Command R+, Anthropic PBC’s Claude 3.5 Sonnet, OpenAI’s GPT-4o and Mistral AI’s Codestral.