v0.1.78

Changelog

Features

• 3c9cb61 feat(anthropic): support Claude subscription tokens and native /v1/messages forwarding (#706)
• 5564295 feat(dashboard): add scope tabs to the rate limits page (#700)
• a2115eb feat(dashboard): fill loading-state gaps and speed up audit log search (#699)
• c70f9d9 feat(providers): add Hetzner experimental inference provider (#701)
• bd0c9c4 feat(providers): add dedicated llamacpp provider type (#698)
• 02636c9 feat(providers): support gemini-3.7-flash and grok-4.6 (#695)

Bug Fixes

• ef1dff0 fix(anthropic): keep system-role messages and their cache breakpoints in place (#713)
• 10caa43 fix(dashboard): Interactions preview the request step sent to the provider (#710)
• ae7556f fix(providers): hide offline providers' models from model listings (#717)
• 33c1578 fix(server): record usage and feedback for non-streaming passthrough responses (#709)

Performance

• ca0d93d perf(responses): move response snapshot writes off the request path (#711)
• 0a91cc2 perf(responsestore): serialize response snapshots once (#715)
• e05466c perf(server): cut hot-path overhead and guard the production-shaped path (#694)