Auto-generated changes

Author: zen-weaver-of-integration[bot]

.changeset/analyze-mem-no-double-count.md

@@ -1,5 +1,22 @@
 # @platforma-open/milaboratories.mixcr-clonotyping.workflow
 
+## 3.28.4
+
+### Patch Changes
+
+- 86d727d: Analyze counts input size once when sizing its memory request, cutting a 110 GiB
+  preset with 31.75 GiB of reads from 237 GiB to 127 GiB.
+
+  The per-analysis baseline — 64 GiB, or 110 and 192 GiB for contig/cell and MiTool
+  presets — is already a total-memory value for the run, so adding a size-derived term
+  on top of it counted the input twice. The request is now
+  `clamp(4 x size, baseMemGiB, 256 GiB)`, making the baseline the lower bound it was
+  always meant to be. Small inputs still receive the full baseline, and both the
+  256 GiB cap and the "Advanced Settings" memory override behave as before.
+
+  The analyze template's `hash_override` UUID also changes, which forces a one-time
+  recompute of cached analyze results.
+
 ## 3.28.3
 
 ### Patch Changes

workflow/CHANGELOG.md

@@ -1,6 +1,6 @@
 {
   "name": "@platforma-open/milaboratories.mixcr-clonotyping-2.workflow",
-  "version": "3.28.3",
+  "version": "3.28.4",
   "private": true,
   "description": "Tengo-based template",
   "type": "module",