Configure optional text-only Ollama model for Jellyfin visual overviews

This commit is contained in:
Sleeps
2026-10-08 15:45:45 +01:00
parent 565fe7520f
commit 8817e483a7
+5
View File
@@ -15,6 +15,11 @@ CAPTION_SPECIALIST_MODEL=aha2025/llama-joycaption-beta-one-hf-llava:Q4_K_M
CAPTION_FORCE_SPECIALIST=false
CAPTION_TIMEOUT_SECONDS=180
CAPTION_MAX_TOKENS=120
# Optional text-only synthesis of stored Trickplay captions into an AI
# visual overview. Never writes to Jellyfin without an explicit CLI --apply.
# Leave empty to use the general caption model (may be a VL model).
OVERVIEW_MODEL=
OVERVIEW_TIMEOUT_SECONDS=240
# Push new caption metadata to Meilisearch in small incremental batches.
CAPTION_MEILI_BATCH_SIZE=500
# Caption adjacent Trickplay frames as chronological windows rather than one