Repository Issue Activity (beta)

#2976 [Epic] Enable safe cross-model KV-cache reuse

#2468 [Security] Constrain OpenClaw image and skill provisioning inputs

#3682 [Chore] Fix Dashboard AMD logo and docs header links

#3349 [Feature] Export provider prompt-cache read and write token metrics

#3718 [Bug] plugin-request-mutations cannot reliably distinguish header overwrite from append

#3758 [Bug] Kubernetes startup skips global config validation

#3756 [Bug] Generic classifier signal's local backend cannot load any registered (ModernBERT-family) model

#3751 [Feature] Expose dashboard (port 8700) in Kubernetes deployments

#3753 [Bug] vllm-sr dashboard --target k8s prints a ClusterIP that cannot be opened

#3725 [Bug] Playground feedback should record for all signed-in users, appear on Insights, and not let shared read accounts change routing

#3715 [Bug] Conversation signal reports a cumulative extraction latency for every rule after the first

#3714 [Bug] Signal match and extraction metrics skip recipe scoping at five call sites

#3612 [Feature] Add initial Fireworks AI serverless model mappings

#2928 [Feature] Hallucination detector adopts the shared TokenClassifierBackend span contract

#2922 [Feature] Define the token_spans.v1 contract for PII backends

#3343 [Feature] Reset eligible history safely on topic change

#3339 [Refactor] Route Looper model calls through the shared connector

#2967 [WG] Data Plane & Networking — charter and Lead/Member self-nomination

#3197 [Research] Establish a versioned quality baseline and gap report for built-in Router Models

#3741 [Enhancement] Live model verification success dot is not visually distinct from idle state

#3734 [Bug] Kubernetes reconcile resets explicit `false` and `0` global values to their defaults

#3738 [Bug] External-API RAG keeps only the first query window

#2470 [Feature] Introduce generation-owned construction, reload, and graceful shutdown

#3704 [Feature] Revalidate transformed request demand before backend dispatch

#3728 [Feature] Return routing latency and cost as response headers