做 AI 应用运维的团队终于可以告别手动修 Bug 了——LangSmith Engine 自动开 PR 和生成评估器，建议直接集成到工作流里试试。
LangSmith Engine 现在能针对每个检测到的问题自动提出三种解决动作：1）自动起草代码或提示词变更并开 PR，供用户审查合并；2）创建针对该问题的自定义在线评估器，防止复发；3）将失败的线上案例加入离线评估数据集，持续提升评估覆盖率。这意味着开发者无需手动排查和修复，系统能主动生成修复方案并扩展测试集，大幅降低 AI 应用维护成本。
For every issue it surfaces, LangSmith Engine proposes three resolution actions. 1️⃣Opens a PR Draf...
For every issue it surfaces, LangSmith Engine proposes three resolution actions. 1️⃣Opens a PR Drafts a targeted code or prompt change + opens against repo. You can review & merge. 2️⃣Creates a custom online evaluator Proposes an evaluator scoped to the exact problem. If it happens again, it gets resurfaced. 3️⃣Adds to your offline eval suite Pulls failing production races into a dataset of ground truth examples ✨Every issue you resolve improves eval coverage along the way. 💬 2 🔄 0 ❤️ 5 👀 810 📊 3 ⚡