Repository Issue Activity (beta)

#197 Add new benchmark mode to search for peak goodput under an SLO

#1158 Native Multi-LoRA / Multi-Tenant Workload Simulation and Weighted Targeting

#1156 Discuss bilingual (English/Chinese) documentation support

#653 Eval Hub integration via SDK

#968 Implement OTEL trace format replay

#1152 Ensure requests are forcefully killed a certain amount of time after benchmark completes.

#1078 Token thoughput and latency metrics do not proper occluded events in warmup

#1145 `inter_token_latency_ms` is a per-request average, not an inter-token latency — the per-interval data is never recorded

#1144 The metric labelled TPOT in CSV / console / plot / HTML is `time_per_output_token_ms`, whose formula matches no other benchmarking tool

#1147 Fix dropped modalities in client, when passed synthetic_text data kind

#1114 Trace hash_ids are not properly scoped

#704 Display benchmark progress in non-interactive consoles

#1113 Running a large mooncake dataset leads to stall

#179 Token Efficiency Benchmarking

#1082 Add conversation-level concurrency for realistic multi-turn user and agent workloads

#582 Mock Server support for multimodal

#1125 OMNI TTS

#1124 vLLM OMNI support

#303 html reporter stopped displaying data

#1024 Support loading pre-generated multi-turn conversation datasets from file

#1104 Support inline and file-based API keys with global per-request rotation

#1083 vLLM Python backends start up twice

#518 Get progress metrics in a file during run, not after the result is completed

#992 Implement full subagent WEKA trace replay

#574 Alerting on internal scheduling problems