Enhancement: Improve matching for correspondents, storage path and labels by removing bias + adding minimum match threshold (#12164)

This commit is contained in:
Philipp Defner
2026-09-09 22:40:24 +00:00
committed by GitHub
parent 310628699d
commit 3e56dace73
4 changed files with 227 additions and 28 deletions
+7
View File
@@ -96,6 +96,13 @@ MODEL_FILE = get_path_from_env(
"PAPERLESS_MODEL_FILE",
DATA_DIR / "classification_model.pickle",
)
# Minimum confidence (0.0-1.0) for the ML classifier to assign a correspondent,
# document type, or storage path. 0.0 disables the threshold.
CLASSIFIER_MATCH_THRESHOLD: Final[float] = get_float_from_env(
"PAPERLESS_CLASSIFIER_MATCH_THRESHOLD",
0.6,
)
LLM_INDEX_DIR = DATA_DIR / "llm_index"
LLM_INDEX_LOCK = LLM_INDEX_DIR / "index.lock"
# Cross-process read/write lock guarding the LLM index compaction/migration