From a822aa5a0efd2300591bbf178697a293312e7276 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Marc-Andr=C3=A9=20Bombeck?= <mbombeck@gmail.com>
Date: Mon, 2 Nov 2026 20:31:33 +0100
Subject: [PATCH] fix(documents): send a read that states no unit itself to
 review

A unit of "1" is the explicit no-unit choice a person makes. When a
document read returns it on its own, the value is staged for review
instead of passing straight through, so a unit the read dropped never
lands as a dimensionless value.
---
 src/lib/documents/__tests__/extract.test.ts | 28 +++++++++++++++++++++
 src/lib/documents/extract.ts                |  8 +++++-
 2 files changed, 35 insertions(+), 1 deletion(-)

diff --git a/src/lib/documents/__tests__/extract.test.ts b/src/lib/documents/__tests__/extract.test.ts
index a5900338c..a637df428 100644
--- a/src/lib/documents/__tests__/extract.test.ts
+++ b/src/lib/documents/__tests__/extract.test.ts
@@ -165,6 +165,34 @@ describe("runInboundExtraction — stated-status-only mapping", () => {
     expect(med.statusStated).toBe("ongoing");
   });
 
+  it('sends a confident value the read itself gave no unit ("1") to review', async () => {
+    const provider = fakeProvider(
+      JSON.stringify({
+        facts: [
+          {
+            type: "OBSERVATION",
+            label: "CHOL/HDLC RATIO",
+            value: 4,
+            unit: "1",
+            confidence: 0.95,
+            sourceText: "CHOL/HDLC RATIO 4.0",
+          },
+          {
+            type: "OBSERVATION",
+            label: "HDL",
+            value: 52,
+            unit: "mg/dL",
+            confidence: 0.95,
+            sourceText: "HDL 52 mg/dL",
+          },
+        ],
+      }),
+    );
+
+    const result = await runInboundExtraction(oneRead(provider));
+    expect(result.facts.map((f) => f.needsReview)).toEqual([true, false]);
+  });
+
   it("stores the document's real span, not the model's echo of it", async () => {
     // The OCR text says "Hämoglobin    14.2 g/dL"; the model echoes a tidied
     // rendering of it. The stored provenance must be the document's characters.
diff --git a/src/lib/documents/extract.ts b/src/lib/documents/extract.ts
index a1660adf5..a21f8eb44 100644
--- a/src/lib/documents/extract.ts
+++ b/src/lib/documents/extract.ts
@@ -57,6 +57,7 @@ import {
   type OcrMeter,
   type OcrReadSurface,
 } from "@/lib/documents/ocr-read";
+import { isDimensionlessUnit } from "@/lib/labs/dimensionless";
 import { annotate } from "@/lib/logging/context";
 import {
   INBOUND_CONFIDENCE_FLOOR,
@@ -383,7 +384,12 @@ export async function runInboundExtraction(
     facts.push({
       factType: raw.type,
       confidence,
-      needsReview: confidence < INBOUND_CONFIDENCE_FLOOR,
+      // A unit of "1" is the user's explicit no-unit choice. A read that
+      // states it itself goes to review rather than straight to the record,
+      // so a dropped unit never passes as a dimensionless value.
+      needsReview:
+        confidence < INBOUND_CONFIDENCE_FLOOR ||
+        (raw.type === "OBSERVATION" && isDimensionlessUnit(raw.unit)),
       data,
       provenance: {
         sourceText: anchor.sourceText,