drjones/draco
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

vornith stability: auto-retry+reload, k=2 prompt budget, budget-sized warm ping, web auto-retry; revert keep_alive

Browse Source
This commit is contained in:
drjones
2026-10-01 18:22:45 -07:00
parent a4ef81a5d6
commit 9e850aabd6
7 changed files with 124 additions and 46 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
deploy/draco-warm.service Normal file
View File

@@ -0,0 +1,6 @@
[Unit]
Description=DRACO model keep-warm ping
[Service]
Type=oneshot
ExecStart=/usr/bin/curl -s -m 130 http://127.0.0.1:8012/api/warm

9
deploy/draco-warm.timer Normal file
View File

@@ -0,0 +1,9 @@
[Unit]
Description=DRACO keep model warm every 4 minutes
[Timer]
OnBootSec=60
OnUnitActiveSec=240
[Install]
WantedBy=timers.target

2
deploy/draco.service
View File

@@ -5,7 +5,7 @@ After=network.target
[Service]
Type=simple
WorkingDirectory=/opt/draco
ExecStart=/opt/draco/venv/bin/gunicorn -w 2 --threads 8 -b 127.0.0.1:8012 --timeout 300 app:app
ExecStart=/opt/draco/venv/bin/gunicorn -w 2 --threads 8 -b 127.0.0.1:8012 --timeout 300 --graceful-timeout 10 app:app
Restart=always
RestartSec=5
Environment=PYTHONUNBUFFERED=1