aboutsummaryrefslogtreecommitdiff
path: root/app/backends/sglomni
AgeCommit message (Collapse)Author
2026-09-09fix: model-specific instructions support and set cfg-scale 4 for breeze ↵historia
automatically
2026-09-04fix: sglomni retry contract, booting detection, CLI keys, companion refresh, ↵historia
ffmpeg dep
2026-09-03fix: increase gpu memory for MOSS with gplang-omni to .87historia
2026-09-03feat: gplang-omni reserves 70% of gpu 0 memory, preventing oom errors on ↵historia
consumer cards
2026-09-02fix: warn before using a likely too-big chunk size for sglang-omni modelshistoria
2026-09-02feat: reserve 20% vram when running higgs with sglang-omnihistoria
2026-09-02feat: sglang-omni model picker on manual server launchhistoria
2026-09-02fix: ZONOS2 only generating 1024 tokens maxhistoria
2026-09-02feat: sglang backend supporthistoria