[Bug]: `redis-semantic` cache never produces semantic hits, `_get_cache_key_filter_expression` uses full request hash as RediSearch pre-filt
![[Bug]: `redis-semantic` cache never produces semantic hits, `_get_cache_key_filter_expression` uses full request hash as RediSearch pre-filt](https://www.chat-gpts.plus/wp-content/uploads/2026/06/29086-12748e58-768x403.jpg)
用户在 LiteLLM 代理中配置了 cache_params.type: redis-semantic ,期望基于语义相似性(向量距离)命中缓存。但实际上,即使两个请求的语义高度相似(余弦距离约 0.008),缓存查找仍然返回 miss,请求被转发到模型并再次写入缓存。该问题在 LiteLLM 版
![[Bug]: GLM-5 tool calls in stream mode get error tool name](https://www.chat-gpts.plus/wp-content/uploads/2026/06/39757-790e149e-768x403.jpg)


![ValueError: [Errno 22] Invalid argument; last error log: [pad] Failed to configure input pad on pad](https://www.chat-gpts.plus/wp-content/uploads/2026/06/14483-74416a64-768x403.jpg)
![[XPU] GGUF Q6_K dequantization segfault on Intel GPU during model loading](https://www.chat-gpts.plus/wp-content/uploads/2026/06/14515-c603a024-768x403.jpg)


![[Bug]: obj serialization mismatch](https://www.chat-gpts.plus/wp-content/uploads/2026/06/21611-5b2ecf70-768x403.jpg)
