2026-04-30 - 2026-07-30

Overview

0 Active Pull Requests
13 Active Issues
Excluding merges, 1 author has pushed 1 commit to main and 24 commits to all branches. On main, 1 file has changed and there have been 3 additions and 0 deletions.

13 Issues created by 1 user

Opened #1 Phase 1: GPU-Manager — Schritte 0–5 (ComfyUI systemd → mVoice load/unload → Broker MVP → wa.sh migrate → Queue → LRU-Eviction) 2026-05-11 11:18:04 +00:00

Opened #2 Add a generic GPU-lease primitive (acquire/renew/release) for long-running async consumers (ComfyUI/FLUX) 2026-06-07 08:49:29 +00:00

Opened #3 mGPUmanager: ollama not evictable + no game-awareness + OLLAMA_KEEP_ALIVE=24h starve the game/FLUX 2026-06-07 09:14:08 +00:00

Opened #4 Broker reports per-consumer gpu_resident_mib=0 for externally-started consumers → eviction finds "no candidates" and never reclaims their VRAM 2026-06-07 13:12:26 +00:00

Opened #5 Scheduler must evict ollama for higher-priority image leases (vram_managed guard defeats the manager's purpose) 2026-07-06 10:35:26 +00:00

Opened #6 Broker must be the SOLE VRAM authority — close bypass gaps so out-of-band ollama holds get evicted (imagen starved by 24h keep-alive) 2026-07-10 10:51:06 +00:00

Opened #7 STT big-model via GPU manager: mVoice eager-loads at startup + PWA bypasses the broker → big model OOM-crashes on a full GPU 2026-07-14 15:29:30 +00:00

Opened #8 unload(): whisper-server systemd branch marks consumer unloaded without stopping the unit (latent VRAM-not-freed bug) 2026-07-14 16:14:54 +00:00

Opened #9 STT forces German on every utterance: -l de hardcoded + language never forwarded, so English voice is translated to German 2026-07-25 09:30:31 +00:00

Opened #10 STT glossary drift: consumers.yaml stt_prompt is hand-maintained and diverges from mVoice lexicon.yaml (missing mBrian → 'mBryan') 2026-07-25 09:30:31 +00:00

Opened #11 comfyui's vram_budget_mib is 11000, measured usage is 4346 — the broker refuses image jobs that would fit comfortably 2026-07-28 23:46:53 +00:00

Opened #12 comfyui's VRAM need is not a constant: it varies cold-vs-warm (8410 vs 6170 MiB) and with the requested resolution 2026-07-29 10:06:38 +00:00

Opened #13 The broker can reclaim VRAM it did not allocate, but nothing stops a direct client taking it — a bare ollama run still OOM-kills whisper 2026-07-29 10:34:34 +00:00