{"id":13777,"date":"2025-06-06T13:28:00","date_gmt":"2025-06-06T21:28:00","guid":{"rendered":"https:\/\/live-cometml.pantheonsite.io\/?p=13777"},"modified":"2025-06-12T16:16:11","modified_gmt":"2025-06-12T16:16:11","slug":"comet-product-releases-june2025","status":"publish","type":"post","link":"https:\/\/www.comet.com\/site\/blog\/comet-product-releases-june2025\/","title":{"rendered":"Release Highlights: Discover Opik Agent Optimizer, Guardrails, & New Integrations"},"content":{"rendered":"\n

As LLMs power more complex, multi-step agentic systems, the need for precise optimization and control is growing. In case you missed the announcement, two new powerful features are now available in Opik to help you build with confidence: Opik\u2019s Agent Optimizer<\/a> for automating prompt and agent tuning, and Guardrails<\/a> for real-time output validation. We invite you to try them out completely free as part of your LLM evaluation<\/a> workflows.<\/p>\n\n\n\n

We\u2019ve also expanded our integration support with LangchainJS<\/a>, Autogen<\/a>, and Agno<\/a>\u2014bringing full tracing and evaluation to the frameworks you rely on across JavaScript, enterprise, and open-source stacks. Read on to learn more!<\/p>\n\n\n\n

Opik LLM Eval Updates<\/h2>\n\n\n\n

We kicked off summer 2025 with a pair of major additions to your LLM application and agent dev cycle. Automated prompt engineering<\/a> helps you act on your eval metrics, iterating to determine the best system prompts for your needs in a fraction of the time this would take to do manually. And with trust and safety at the forefront of your business’s brand and reputation, built-in guardrails<\/a> add much needed peace of mind when shipping user-facing LLM applications to production.<\/p>\n\n\n\n

Agent Optimizer SDK<\/h3>\n\n\n\n
\"\"<\/figure>\n\n\n\n

<\/p>\n\n\n\n

Opik Agent Optimizer<\/a> is a new SDK in public beta that brings automated prompt and agent optimization to your LLM development workflows. It supports multi-trial search and integrates with Opik\u2019s evaluation stack, helping you measure, compare, and refine agent behavior at scale. The SDK includes a suite of optimization strategies, such as meta-prompting, few-shot example tuning, and evolutionary search, and can be used across a wide range of models and frameworks. With just a few lines of code, you can begin systematically optimizing your prompts and agents using real evaluation data.<\/p>\n\n\n\n

Agent Optimizer Docs<\/a><\/p>\n\n\n\n

Guardrails<\/h3>\n\n\n\n
\"\"<\/figure>\n\n\n\n

<\/p>\n\n\n\n

Guardrails<\/a> is a new feature that enables real-time screening and control of LLM application behavior. It provides a lightweight validation layer for detecting and managing undesired outputs, such as off-topic responses, policy violations, or sensitive data exposure, before they reach users. Guardrails includes native support for topic and PII checks, and can be extended with custom or third-party moderation logic. Designed to integrate directly into agent workflows, it helps you to enforce output standards and deploy with greater confidence.<\/p>\n\n\n\n

Guardrails Docs<\/a><\/p>\n\n\n\n

Opik Integrations<\/h2>\n\n\n\n

Opik is designed to fit seamlessly into your existing AI engineering stack, and we’re constantly shipping new integrations to support this. Here are three of the latest:<\/p>\n\n\n\n

LangchainJS<\/h3>\n\n\n\n

We’re excited to support LangchainJS, the JavaScript\/TypeScript version of one of the most popular building blocks for LLM applications and agents. This integration makes it easy to trace and evaluate LangchainJS pipelines directly in Opik, helping to bring observability to your agent workflows.<\/p>\n\n\n\n

Opik & LangchainJS Integration Docs<\/a><\/p>\n\n\n\n

Autogen<\/h3>\n\n\n\n

Opik now supports Autogen, Microsoft\u2019s enterprise-focused agent framework designed for orchestrating complex multi-agent interactions. This integration provides deep trace visibility into Autogen workflows, making it easier for you to debug, evaluate, and scale high-performance agents.<\/p>\n\n\n\n

Opik & Autogen Integration Docs<\/a><\/p>\n\n\n\n

Agno<\/h3>\n\n\n\n

Agno is one of the fastest-growing open-source agent frameworks, and now you can trace and evaluate Agno agents natively in Opik. If you are a developer working in community-led stacks, this integration gives you full access to Opik\u2019s observability tools, enabling smarter iteration and more reliable agent behavior.<\/p>\n\n\n\n

Opik & Agno Integration Docs<\/a><\/p>\n\n\n\n

Additional Opik Releases<\/h2>\n\n\n\n