模型行为官方发布自动监测

LiteLLM v1.104.4

LiteLLM v1.104.4 is a patch release on the stable/1.104.x branch that backports fixes for guardrails, proxy team-scoped batch model access, and token counting.

原始内容为英文;当前页面提供中文导航与来源说明,具体事实请以原文为准。

人类阅读

为什么值得关注

LiteLLM released patch version v1.104.4 for the stable/1.104.x branch. Changes include: restoring supported decisions providers for loading guardrail settings, backporting team-scoped batch model access to the proxy, backporting October 19 token counting fixes, restoring Google request compatibility in count tokens, and restoring legacy token count input precedence. Docker container images are signed with cosign and can be verified against the official public key.

Agent 解析

可执行摘要

LiteLLM release v1.104.4 backports multiple bug fixes to stable/1.104.x, addressing guardrail provider loading, team-scoped batch model access, Google request compatibility in token counting, and legacy token count input precedence. Docker images are signed with cosign.

Agent 实用度
62/100
可信度
96%
机器格式
JSON + Markdown
下一步

开发者应核对什么

  • Upgrade to v1.104.4 if running on the stable/1.104.x branch and experiencing issues with guardrail settings, proxy batch model access, or token counting.
  • Verify Docker container integrity using cosign with the published public key if pulling ghcr.io/berriai/litellm:v1.104.4.
分类

标签与路由

gatewaymodelsobservability
相关信号

继续阅读