Repository Issue Activity (beta)

#107 Feature request: deterministic verbatim utterance mode for fixed announcements

#106 Service-Duplex-Bench

#105 --cpu-offload crashes at warmup: KV cache allocated on meta device for offloaded layers ("Cannot copy out of meta tensor")

#96 DGX Spark / GB10: PersonaPlex runs but local Web UI audio latency continuously increases

#3 PersonaPlex produces choppy/unusable audio on DGX Spark (GB10)

#71 MLX support for Apple Silicon

#99 No Retention of system prompt post the context window

#97 The model cannot distinguish its role

#93 Tool call support?

#95 Feature request: typed user turns mid-session on /api/chat

#94 Smaller Checkpoint

#66 Connecting RAG and tool calls

#67 [Question] Details on negative-duration silence for barge-in/interruption simulation in training data

#26 Can we train it on other languages?

#11 Instructions for Mac?

#85 Speaker option for this model

#51 Floating Point 0 Error With H20

#83 Bug found during model execution on Linux with CUDA (PyTorch backend)

#73 Spanish from Spain are too people

#38 Multi-GPU support

#74 remove israel name from the readme page

#46 Clarification needed: Purpose of duplicate MimiModel instance (other_mimi) in server.py

#68 Voice clone

#65 RAG data

#56 I have done my due diligence in trying to find the answer myself.