Add distribution aggregation and dual-write

bbconf/db_utils.py

@@ -14,7 +14,7 @@
     event,
     select,
 )
-from sqlalchemy.dialects.postgresql import ARRAY, JSON
+from sqlalchemy.dialects.postgresql import ARRAY, JSON, JSONB
 from sqlalchemy.engine import URL, Engine, create_engine
 from sqlalchemy.event import listens_for
 from sqlalchemy.exc import IntegrityError, ProgrammingError
@@ -253,10 +253,10 @@ class BedStats(Base):
     intron_percentage: Mapped[Optional[float]]
     intergenic_percentage: Mapped[Optional[float]]
     promotercore_percentage: Mapped[Optional[float]]
-    tssdist: Mapped[Optional[float]]
+    median_neighbor_distance: Mapped[Optional[float]]
 
     distributions: Mapped[Optional[dict]] = mapped_column(
-        JSON,
+        JSONB,
         nullable=True,
         comment="Full distribution arrays from gtars genomicdist (JSONB)",
     )
@@ -344,7 +344,7 @@ class BedSets(Base):
         JSON, comment="Median values of the bedset"
     )
     bedset_stats: Mapped[Optional[dict]] = mapped_column(
-        JSON,
+        JSONB,
         nullable=True,
         comment="Pre-aggregated distribution statistics from gtars (JSONB)",
     )

bbconf/models/bed_models.py

@@ -51,6 +51,7 @@ class BedStatsModel(BaseModel):
     number_of_regions: float | None = None
     gc_content: float | None = None
     median_tss_dist: float | None = None
+    median_neighbor_distance: float | None = None
     mean_region_width: float | None = None
 
     exon_frequency: float | None = None
@@ -209,6 +210,13 @@ class BedListResult(BaseModel):
     results: list[BedMetadataBasic]
 
 
+class BedBatchResult(BaseModel):
+    count: int
+    limit: int
+    offset: int
+    results: list[BedMetadataAll]
+
+
 class QdrantSearchResult(BaseModel):
     id: str
     payload: dict = None

bbconf/models/bedset_models.py

@@ -22,18 +22,31 @@ class BedSetDistributions(BaseModel):
 
     Stored in the bedset_stats JSONB database column. Populated when
     member bed files have been processed with the gtars analysis backend.
+
+    Only distributions that are meaningful at collection-level are kept:
+    - scalar_summaries: mean ± sd + 25-bin histograms of per-file scalar values
+    - tss_histogram: summed per-bin counts across files (fixed ±100 kb axis)
+    - region_distribution: per-chrom bin-wise mean ± sd across files
+      (requires gtars ≥ PR #248 for reference-aligned bin widths)
+    - partitions: mean ± sd of per-file partition percentages
+
+    Dropped (retained in per-file distributions blob, not aggregated):
+    - widths_histogram: per-file variable-range bins aren't summable; use
+      scalar_summaries.mean_region_width histogram instead
+    - neighbor_distances KDE: per-file within-bedset variance is low; use
+      scalar_summaries.median_neighbor_distance instead
+    - gc_content KDE: per-file distribution is unimodal; use
+      scalar_summaries.gc_content mean instead
+    - chromosome_summaries: redundant with region_distribution
+    - expected_partitions: per-file null hypothesis, not collection property
     """
 
     n_files: int = 0
     composition: Optional[dict] = None
     scalar_summaries: Optional[dict] = None
     tss_histogram: Optional[dict] = None
-    widths_histogram: Optional[dict] = None
-    neighbor_distances: Optional[dict] = None
-    gc_content: Optional[dict] = None
     region_distribution: Optional[dict] = None
     partitions: Optional[dict] = None
-    chromosome_summaries: Optional[dict] = None
 
 
 class BedSetPlots(BaseModel):