发布

  • [NA] [Docs] Update changelog (#3929)

    frostbyte_neo 发布于 2025-11-04 11:03:36 +00:00

    This changelog entry covers releases 1.8.83-1.8.97 (November 4, 2024) and highlights the following major improvements:

    🚨 Native Slack and PagerDuty Alerts

    • Added native integrations for Slack and PagerDuty, eliminating middleware requirements
    • Direct configuration from Opik settings with automatic message formatting

    🖼️ Multimodal LLM-as-a-Judge Support

    • Added visual evaluation capabilities for traces containing images
    • Support for vision-capable models to evaluate image generation quality and multimodal content

    Prompt Generator & Improver

    • New AI-powered tools for creating and refining prompts
    • Available directly in the Playground with best practices from OpenAI, Anthropic, and Google

    🔗 Advanced Prompt Integration

    • Seamless connection between Prompt Library, Traces, and Playground
    • Associate prompts with traces and spans using opik_context module

    🤖 Agent Evaluation Guides

    • New comprehensive guides for evaluating agent trajectories and multi-turn agents
    • Includes diagrams and step-by-step instructions

    🧪 No-Code Experiment Capabilities

    • Enhanced Playground with dataset creation, online score rules, and experiment navigation
    • Reference-based experiments with automatic comparison to expected answers

    📊 On-Demand Online Evaluation

    • Run metrics on existing traces and threads without code
    • Perfect for evaluating historical data or backfilling new scores

    📦 Import/Export Commands

    • New CLI commands for data migration, backup, and version control
    • Export/import traces, spans, datasets, prompts, and evaluation rules
    下载附件