Repository Issue Activity (beta)

#2606 internal error PoolTimeout leads job failure

#238 Should ModelAdapter displays the `Status` on the result of get command

#1416 [RFC]: AIBrix Documentation and samples improvement

#2605 PD disaggregation routing: observed cross-roleset decode pod selection with concurrent multi-model workloads (not reproducible on current nightly)

#2596 [Docs] Add missing gateway target and general headers

#2552 active_jobs has some stale job information which maybe done

#2591 KPA/APA should not scale down on unknown metrics and should reject invalid subTargetSelector targets

#2583 RoleSet and StormService should reject invalid role specs during admission

#2561 Discussion: Cache-aware rescheduling for recreated StormService/RoleSet pods

#2578 Gateway double-wraps upstream error responses from vLLM/SGLang

#2597 [Cleanup] Avoid duplicated session affinity header constants

#2346 v0.8.0 Roadmap

#2568 Request-aware automatic config profile selection

#2574 [RFC]: Enable Heterogeneous SLO Routing & GPU Optimizer for Ascend NPUs (910B / 910C)

#2590 PodAutoscaler HPA strategy accepts invalid replica bounds and inconsistent targetValue formats

#2579 [Feature] Model Scaling and Request Routing for Multi-LoRA Serving

#2285 KV event decoder drops group_idx/medium/lora_name, breaking prefix matching for hybrid and multi-tier models

#2520 Support scheduled replica bounds for PodAutoscaler

#2573 [Bug] LocalStorage atomic write fails intermittently on Windows when a reader holds the destination

#2546 [Bug] Batch inference retries default to 120, letting a single request retry for ~70 minutes and outlive its request lock

#2569 Gateway cache metrics collection is too aggressive and lacks freshness semantics

#2540 [Bug] Benchmark connection timeout when creating or receiving chunks

#2565 [Bug] S3/TOS downloaders ignore the per-request force_download override

#2560 Discussion: First-class image warmup workflow for AIBrix model workloads

#2537 [Bug] RoleSet scale-down gets stuck and leaves Pending (not-ready) pods when ready pods are at the maxUnavailable floor