Arthur Erlich

haylan commented on issue haylan/LLM-Server#41

2026-09-05 20:42:02 +00:00

Add comfyui service to docker-compose.yml, register with OmniRoute

Compose scaffolding done and pushed (ed83fca): comfyui service (yurisasc/comfyui-rocm7.1, gfx1201-tuned), ai-stack network, host port 8138 published, GID/UID auto-resolved in scripts/update.sh.

Re…

haylan pushed to main at haylan/LLM-Server

2026-09-05 20:40:10 +00:00

ed83fca05c feat(comfyui): add local image-gen service (#41)

haylan pushed to main at haylan/LLM-Server

2026-09-05 19:24:25 +00:00

7d1ff2f54f docs(research): add image-gen backend survey and omniroute/qwen websearch notes

haylan closed issue haylan/LLM-Server#39

2026-09-05 19:16:32 +00:00

Best diffusion model given full ~32GB headroom (Qwen stopped)

haylan commented on issue haylan/LLM-Server#39

2026-09-05 19:16:31 +00:00

Best diffusion model given full ~32GB headroom (Qwen stopped)

Recommend Qwen-Image at FP8 precision (~25GB: fp8 diffusion weights + fp8 text encoder) as the upgrade from FLUX.1-schnell now that the full ~32GB is available. It has the cleanest license…

haylan pushed to main at haylan/LLM-Server

2026-09-05 19:03:19 +00:00

ac3f730f83 docs(research): recommend Qwen-Image FP8 for full-VRAM diffusion build

haylan closed issue haylan/LLM-Server#40

2026-09-05 18:57:39 +00:00

Why lazytainer idle-stop doesn't work today, and how the switch script should handle it

haylan commented on issue haylan/LLM-Server#40

2026-09-05 18:57:36 +00:00

Why lazytainer idle-stop doesn't work today, and how the switch script should handle it

Confirmed: lazytainer's packet-threshold detector (gopacket/libpcap BPF filter on the port, counting every packet, no source/traffic-class filtering — confirmed in vmorganp/Lazytainer source)…

haylan pushed to main at haylan/LLM-Server

2026-09-05 18:57:28 +00:00

451d5c7b28 docs(research): confirm lazytainer/omniroute idle-stop conflict (#40)

haylan opened issue haylan/LLM-Server#43

2026-09-05 17:36:39 +00:00

scripts/switch-model.sh: swap GPU residency between Qwen and ComfyUI

haylan opened issue haylan/LLM-Server#42

2026-09-05 17:36:38 +00:00

Extend downloader for the diffusion model weights

haylan opened issue haylan/LLM-Server#41

2026-09-05 17:36:37 +00:00

Add comfyui service to docker-compose.yml, register with OmniRoute

haylan opened issue haylan/LLM-Server#40

2026-09-05 17:36:36 +00:00

Why lazytainer idle-stop doesn't work today, and how the switch script should handle it

haylan opened issue haylan/LLM-Server#39

2026-09-05 17:36:34 +00:00

Best diffusion model given full ~32GB headroom (Qwen stopped)

haylan opened issue haylan/LLM-Server#38

2026-09-05 17:36:14 +00:00

Local image generation via ComfyUI + OmniRoute

haylan pushed to main at haylan/LLM-Server

2026-09-05 10:00:11 +00:00

d8736b6dd7 fix(llama-server): quantize KV cache, restore full 262144 context

haylan pushed to main at haylan/LLM-Server

2026-09-05 09:32:21 +00:00

5767f548c3 perf(llama-server): enable flash attention

haylan pushed to main at haylan/LLM-Server

2026-09-05 09:29:16 +00:00

633292b291 fix(omniroute): raise stream-idle timeout to 180s

haylan pushed to main at haylan/LLM-Server

2026-09-05 09:26:39 +00:00

23e90fe8fb fix(llama-server): cap concurrent slots at 2 to curb prefill contention

haylan pushed to main at haylan/LLM-Server

2026-09-04 19:47:45 +00:00

ae812cd9e0 feat(ctx): larger context size