JetBrains has announced a new open-source library called Tracy, designed to help developers trace and monitor AI features in their Kotlin and Java projects. The tool was unveiled on March 11 and is available on GitHub. Tracy aims to make it easier for programmers to track AI model calls, debug issues, and measure performance directly within their code.

What Tracy Does

Tracy provides a unified API that captures structured traces from AI-related activities in Kotlin and Java applications. It can record details like execution times, inputs and outputs of functions, and usage of large language models (LLMs). This makes it easier for developers to identify where problems occur and optimize their AI workflows.

The library supports tracing different aspects of AI integration, such as monitoring messages sent to AI clients, tracking token usage, and measuring costs associated with model calls. It also allows for manual creation and management of spans, giving developers control over how they monitor specific parts of their code.

Compatibility and Integration

Tracy is compatible with Kotlin from version 2.0.0 and Java from version 17. It adopts the OpenTelemetry Generative AI Semantic Conventions, ensuring that traces follow widely accepted standards and can work with any OpenTelemetry-compliant backend. Currently, Tracy can export traces to platforms like Langfuse and Weave, making it easy to visualize and analyze data.

JetBrains also highlighted that Tracy integrates smoothly with popular SDKs from AI providers such as OpenAI, Anthropic, and Gemini. It works well with common Kotlin/LLM stacks, including OkHttp and Ktor clients, making it straightforward for developers to add tracing features to their existing projects.

Overall, Tracy offers a powerful toolset for developers looking to improve the transparency and debugging capabilities of AI features in their Kotlin and Java applications. Its open-source nature and compatibility with standard protocols make it a versatile choice for those working with AI integrations.

Inspired by

• https://www.infoworld.com/article/4144520/jetbrains-unveils-ai-tracing-library-for-kotlin-and-java.html

Sources

• jetbrains.com
• github.com
• opentelemetry.io