Commit graph

3506 commits

Author SHA1 Message Date
Auto-Sync
350d60f7b9 Auto-Sync: 2026-03-26 07:00 2026-03-26 07:00:10 +01:00
Auto-Sync
ebb2931415 Auto-Sync: 2026-03-26 06:45 2026-03-26 06:45:08 +01:00
Auto-Sync
fda4864308 Auto-Sync: 2026-03-26 06:30 2026-03-26 06:30:09 +01:00
Auto-Sync
c08873074f Auto-Sync: 2026-03-26 06:15 2026-03-26 06:15:09 +01:00
Auto-Sync
63201b81d4 Auto-Sync: 2026-03-26 06:00 2026-03-26 06:00:09 +01:00
Auto-Sync
01bec5cba0 Auto-Sync: 2026-03-26 05:45 2026-03-26 05:45:09 +01:00
Auto-Sync
b97bbfc2cb Auto-Sync: 2026-03-26 05:30 2026-03-26 05:30:10 +01:00
Auto-Sync
23bc18ce01 Auto-Sync: 2026-03-26 05:15 2026-03-26 05:15:09 +01:00
Auto-Sync
a16d907c53 Auto-Sync: 2026-03-26 05:00 2026-03-26 05:00:09 +01:00
Auto-Sync
efdc587dea Auto-Sync: 2026-03-26 04:45 2026-03-26 04:45:09 +01:00
Auto-Sync
1c49a3aa42 Auto-Sync: 2026-03-26 04:30 2026-03-26 04:30:09 +01:00
Auto-Sync
2cdc22e424 Auto-Sync: 2026-03-26 04:15 2026-03-26 04:15:09 +01:00
Auto-Sync
c04d1cb681 Auto-Sync: 2026-03-26 04:00 2026-03-26 04:00:09 +01:00
Auto-Sync
e8227bc899 Auto-Sync: 2026-03-26 03:45 2026-03-26 03:45:09 +01:00
Auto-Sync
8d200be1a6 Auto-Sync: 2026-03-26 03:30 2026-03-26 03:30:09 +01:00
Auto-Sync
a1f84fc897 Auto-Sync: 2026-03-26 03:15 2026-03-26 03:15:09 +01:00
Auto-Sync
a5e8c4b0fe Auto-Sync: 2026-03-26 03:00 2026-03-26 03:00:09 +01:00
Auto-Sync
0fa3442e76 Auto-Sync: 2026-03-26 02:45 2026-03-26 02:45:09 +01:00
Auto-Sync
4c9ff6dd96 Auto-Sync: 2026-03-26 02:30 2026-03-26 02:30:09 +01:00
Auto-Sync
15c17fed2c Auto-Sync: 2026-03-26 02:15 2026-03-26 02:15:09 +01:00
Auto-Sync
dea7225258 Auto-Sync: 2026-03-26 02:00 2026-03-26 02:00:09 +01:00
Auto-Sync
cf5ac8e22b Auto-Sync: 2026-03-26 01:45 2026-03-26 01:45:09 +01:00
Auto-Sync
e69b51c465 Auto-Sync: 2026-03-26 01:30 2026-03-26 01:30:10 +01:00
Auto-Sync
828190a188 Auto-Sync: 2026-03-26 01:15 2026-03-26 01:15:09 +01:00
Auto-Sync
1e3436e79c Auto-Sync: 2026-03-26 01:00 2026-03-26 01:00:10 +01:00
Auto-Sync
87b3d99471 Auto-Sync: 2026-03-26 00:45 2026-03-26 00:45:09 +01:00
Auto-Sync
a3fa43e9b4 Auto-Sync: 2026-03-26 00:30 2026-03-26 00:30:09 +01:00
Auto-Sync
f169538729 Auto-Sync: 2026-03-26 00:15 2026-03-26 00:15:09 +01:00
Auto-Sync
82b1b1a60a Auto-Sync: 2026-03-26 00:00 2026-03-26 00:00:10 +01:00
Auto-Sync
f8582cee13 Auto-Sync: 2026-03-25 23:45 2026-03-25 23:45:09 +01:00
Auto-Sync
a1216c768f Auto-Sync: 2026-03-25 23:30 2026-03-25 23:30:09 +01:00
Auto-Sync
c867939761 Auto-Sync: 2026-03-25 23:15 2026-03-25 23:15:09 +01:00
Auto-Sync
cfd09c4f23 Auto-Sync: 2026-03-25 23:00 2026-03-25 23:00:09 +01:00
Auto-Sync
af12bbad5b Auto-Sync: 2026-03-25 22:45 2026-03-25 22:45:09 +01:00
Auto-Sync
5f274a1dcc Auto-Sync: 2026-03-25 22:30 2026-03-25 22:30:09 +01:00
Auto-Sync
b4e9dc8fb8 Auto-Sync: 2026-03-25 22:15 2026-03-25 22:15:09 +01:00
Auto-Sync
942f3e032a Auto-Sync: 2026-03-25 22:00 2026-03-25 22:00:09 +01:00
Auto-Sync
ab8fdb67ae Auto-Sync: 2026-03-25 21:45 2026-03-25 21:45:09 +01:00
Auto-Sync
a16b0a7184 Auto-Sync: 2026-03-25 21:30 2026-03-25 21:30:09 +01:00
Auto-Sync
9d5e5564de Auto-Sync: 2026-03-25 21:15 2026-03-25 21:15:09 +01:00
Homelab Cursor
07b785ece1 docs: GPU-Architektur Doku (STATE.md, RAGFLOW.md)
- STATE.md: Hybrid-Architektur, LLM-Routing, VRAM-Belegung
- RAGFLOW.md: Warnung dass qwen2.5:14b Hauptmodell verdraengt
2026-03-25 21:01:47 +01:00
Auto-Sync
0b99490909 Auto-Sync: 2026-03-25 21:00 2026-03-25 21:00:10 +01:00
Homelab Cursor
d2a6391f52 feat(llm): Ollama warmup bei Start - Modelle permanent im VRAM
- warmup_ollama() laedt qwen3:30b-a3b + nomic-embed-text mit keep_alive=-1
- Wird beim Bot-Start in post_init() aufgerufen (via asyncio.to_thread)
- keep_alive=-1 nur ueber native Ollama API (/api/generate) moeglich
- GPU haelt 22.6/24 GB permanent: Text + Embeddings ohne Swap
2026-03-25 20:59:30 +01:00
Homelab Cursor
a0724ba6f1 refactor(llm): GPU-Architektur - Text lokal, Vision Cloud
- MODEL_VISION von qwen3-vl:32b (Ollama) auf openai/gpt-4o-mini (OpenRouter)
- Vision-Modell aus OLLAMA_MODELS entfernt (kein GPU-Swap mehr)
- keep_alive=-1 fuer Ollama: Textmodell bleibt permanent im VRAM
- RTX 3090 wird dedizierter Text+Embedding-Server
- Neue Dienste koennen Ollama mitnutzen ohne Konkurrenz
2026-03-25 20:56:55 +01:00
Auto-Sync
5b57bfd27f Auto-Sync: 2026-03-25 20:45 2026-03-25 20:45:09 +01:00
Auto-Sync
9121e3df3b Auto-Sync: 2026-03-25 20:30 2026-03-25 20:30:09 +01:00
Auto-Sync
314457a513 Auto-Sync: 2026-03-25 20:15 2026-03-25 20:15:09 +01:00
Auto-Sync
6137437d49 Auto-Sync: 2026-03-25 20:00 2026-03-25 20:00:10 +01:00
Homelab Cursor
92898a33e3 fix(llm): kein Fallback auf Textmodell bei Vision-Anfragen
Wenn qwen3-vl:32b timeout hat, wurde bisher auf qwen2.5:14b
(Textmodell) zurueckgefallen. Das kann keine Bilder sehen und
halluziniert stattdessen. Jetzt: allow_fallback=False fuer Vision
und klare Fehlermeldung bei Timeout.
2026-03-25 19:46:35 +01:00
Auto-Sync
ac0f6fda2e Auto-Sync: 2026-03-25 19:45 2026-03-25 19:45:09 +01:00