Skip to content

~95% on SimpleQA (e.g. Qwen3.6-27B on a 3090). Supports all local and cloud LLMs (llama.cpp, Ollama, Google, ...). 10+ search engines - arXiv, PubMed, your private documents. Everything Local & Encrypted.

fix(library): project ResearchResource whole-table scans to bound memory (#4560) (#4724)

LearningCircuit8c4029204455
diff --git a/changelog.d/+bound-research-resource-loads.bugfix.md b/changelog.d/+bound-research-resource-loads.bugfix.md
new file mode 100644
index 0000000000..9b25a35591
--- /dev/null
+++ b/changelog.d/+bound-research-resource-loads.bugfix.md
@@ -0,0 +1 @@
+The "download all" and "queue all undownloaded" library actions now project only the columns they use instead of loading full `ResearchResource` rows, avoiding materializing every resource's `content_preview` text on a whole-table scan for large libraries (#4560).
diff --git a/src/local_deep_research/research_library/routes/library_routes.py b/src/local_deep_research/research_library/routes/library_routes.py
index f01b870150..168bea8ff2 100644
--- a/src/local_deep_research/research_library/routes/library_routes.py
+++ b/src/local_deep_research/research_library/routes/library_routes.py
@@ -615,8 +615,15 @@ def download_all_text():
try:
# Get all undownloaded resources
with get_user_db_session(username) as session:
- # Get resources that don't have text files yet
- all_resources = session.query(ResearchResource).all()
+ # Project only the columns the loop uses (id/url/title) rather
+ # than loading full ``ResearchResource`` entities — the
+ # ``content_preview`` Text column would otherwise materialize
+ # for every resource on this whole-table scan (#4560).
+ all_resources = session.query(
+ ResearchResource.id,
+ ResearchResource.url,
+ ResearchResource.title,
+ ).all()
# Filter to only downloadable resources (academic/PDF)
resources = [
r for r in all_resources if is_downloadable_url(r.url)
@@ -1027,9 +1034,16 @@ def queue_all_undownloaded():
logger.info(f"queue_all_undownloaded called for user {username}")
with get_user_db_session(username) as db_session:
- # Find all resources that don't have a completed download
+ # Find all resources that don't have a completed download. Project only
+ # the columns the filter/loop use (id/url/research_id) instead of full
+ # ``ResearchResource`` entities so the ``content_preview`` Text column
+ # isn't materialized for every row on this whole-table scan (#4560).
undownloaded = (
- db_session.query(ResearchResource)
+ db_session.query(
+ ResearchResource.id,
+ ResearchResource.url,
+ ResearchResource.research_id,
+ )
.outerjoin(
Document,
(