JetBrains has introduced the launch of its new AI coding agent, Junie. Junie runs inside JetBrains IDEs and may tackle easy coding duties or help on extra complicated duties with collaboration from the developer.
“Because of the facility of JetBrains IDEs, coupled with dependable LLMs, Junie already solves duties that may in any other case require hours of labor,” Andrew Zakonov, product chief at JetBrains, wrote in a weblog put up.
Builders can share prompts with the agent, evaluation the outcomes, and regulate as wanted. Over time, it learns the context of code and the developer’s preferences and elegance. “This leads to higher code high quality and management on how Junie performs duties, making certain reliability, making Junie a trusted collaborator in your staff,” Zakonov wrote.
RELATED: AI brokers are reworking the software program growth life cycle
It may additionally run code and exams, and may verify the challenge state after making modifications to confirm that each one exams have handed. “AI-generated code could be simply as flawed as developer-written code. In the end, Junie is not going to simply pace up growth – it’s poised to lift the bar for code high quality,” stated Zakonov.
Junie was in a position to resolve 53.6% of duties on a single run in line with the SWEBench Verified benchmark, which incorporates 500 developer duties. In response to JetBrains, this proves that’s able to adapting to the wants of at the moment’s builders.
In response to JetBrains, Junie doesn’t run regionally, because it makes use of OpenAI and Anthropic fashions. There’s presently a waitlist for the Early Entry Program, and it’s initially accessible in IntelliJ IDEA Final and PyCharm Skilled, with assist for WebStorm coming subsequent. It is just accessible on OS X and Linux in the meanwhile.
