Repository Issue Activity (beta)

#5307 BUG: task=parse rejects a 31-page A4 PDF at every zoomin, and lowering zoomin makes it worse

#5494 [Docker CD] Failure on main

#5484 V3.2.0版本，rerank接口调用无响应

#5466 Make runtime model registry refresh idempotent and concurrency-safe

#5410 在官网的内置模型--->大语言模型里面有qwen3.8模型，但是在xinference v3.2.0 docker镜像的webUI里面找不到

#4633 ENH: SDNQ Diffusers Quantizer

#5467 Merge engine capabilities across heterogeneous workers

#5176 Unauthenticated arbitrary-path file read (partial) via /v1/models/llm/auto-register, plus unauthenticated arbitrary module import via flexible launcher

#5482 部署本地嵌入模型jina-embeddings-v3报错400

#5473 [Question] How to increase context size of Qwen3.6 from UI ?

#5302 deploy failed

#5464 Add startup hooks for distribution-specific model engines

#5459 FIX: apply download worker environment to snapshots

#5347 BUG: xinference 3.2.0 wheel is unimportable - KeyError: 'model_specs' (an audio model entry is packaged into llm_family.json)

#5356 Docs: hybrid setups — self-hosted Xinference + AnyLLM API gateway for frontier models

#5441 not enough values to unpack (expected 2, got 1)

#5284 [bug] Windows: log rotation crashes with WinError 32 because fcntl-based cross-process lock is a no-op

#5344 Cannot launch single-file MTP model using llama.cpp engine (Qwen3.6-MTP gguf support request)

#5450 [Docker CD] Failure on v3.3.0

#5382 [Docker CD] Failure on main

#5433 ENH: support Base64 video inputs in vLLM multimodal chat

#5440 'NoneType' object has no attribute 'update_cache_status'

#5314 qwen3.5默认虚拟环境兼容性缺陷

#5428 Autostart-configured model is never relaunched after a whole worker dies (heartbeat timeout / reverse-channel dead)

#5283 [Feature request]Dynamic scaling model replicas.