Merge origin/release/v1.46.0 into feat/v146-ai

src/__tests__/ai-egress-capability-guard.test.ts

@@ -103,20 +103,15 @@ const EGRESS_SITES: Record<string, EgressSite> = {
     reason:
       "Document summary and transcription; reached from the summary and index routes, the summary job, and the background run worker (`lib/jobs/document-ai-run.ts`), which resolve `documentAi` first.",
   },
-  "lib/documents/extract.ts": {
-    recheck: false,
-    reason:
-      "Document extraction/indexing; reached from the document routes and the index jobs, which resolve `documentAi` first.",
-  },
   "lib/medications/ingredients/label-extract.ts": {
     recheck: false,
     reason:
       "Supplement label read (v1.43); reached from the background run worker (`lib/jobs/document-ai-run.ts`), which re-checks `supplementLabel` for the picked provider, for a label the ingredients extract route queued under that capability.",
   },
-  "lib/labs/ocr-extract.ts": {
+  "lib/documents/ocr-read.ts": {
     recheck: true,
     reason:
-      "Lab report scan; reached from the background run worker (`lib/jobs/document-ai-run.ts`), which checks `labsOcr` for the picked provider before the first read. A long report's reads run for many minutes after that, so every further read re-checks the wire (`aiEgressRefusal`) and stops when AI was turned off meanwhile.",
+      "Report transcription, the one read core of the lab scan (`labsOcr`) and the document extraction (`documentAi`, the manual extract and the lab auto-staging); reached from the background run worker, the extract route and the index job, which check the capability for the picked provider before the first read. A long report's reads run for many minutes after that, so every read re-checks the wire (`aiEgressRefusal`, for the surface's own capability) and stops when AI was turned off meanwhile.",
   },
   "app/api/medications/extract/route.ts": {
     recheck: false,

src/__tests__/baselines/source-file-lines.json

@@ -1,7 +1,7 @@
 {
   "$comment": "Production files over 800 lines and their length. Maintained by src/__tests__/source-size-ratchet.test.ts.",
   "count": 89,
-  "total": 99479,
+  "total": 99397,
   "entries": {
     "app/medications/page-client.tsx": 839,
     "components/charts/mood-chart.tsx": 1166,
@@ -88,7 +88,7 @@
     "lib/sharing/grants.ts": 807,
     "lib/support/report.ts": 903,
     "lib/telegram-webhook-handlers.ts": 1341,
-    "lib/validations/inbound-documents.ts": 1130,
+    "lib/validations/inbound-documents.ts": 1048,
     "lib/withings/sync-sleep.ts": 884,
     "lib/withings/sync.ts": 933,
     "proxy.ts": 870

src/__tests__/baselines/source-function-lines.json

@@ -1,7 +1,7 @@
 {
   "$comment": "Production functions over 200 lines, keyed \"file :: name\", and their length. Maintained by src/__tests__/source-size-ratchet.test.ts.",
-  "count": 354,
-  "total": 114410,
+  "count": 352,
+  "total": 113991,
   "entries": {
     "app/achievements/page.tsx :: AchievementsPage": 206,
     "app/api/admin/backups/upload/route.ts :: apiHandler(arg)": 314,
@@ -95,7 +95,7 @@
     "components/documents/document-bulk-bar.tsx :: DocumentBulkBar": 204,
     "components/documents/document-card.tsx :: DocumentCard": 354,
     "components/documents/document-detail-sheet.tsx :: DocumentDetailSheet": 325,
-    "components/documents/document-facts-review.tsx :: DocumentFactsSection": 252,
+    "components/documents/document-facts-review.tsx :: DocumentFactsSection": 246,
     "components/documents/document-facts-review.tsx :: FactRow": 262,
     "components/documents/document-filter-bar.tsx :: DocumentFilterBar": 348,
     "components/documents/document-record-links.tsx :: DocumentRecordLinks": 202,
@@ -323,8 +323,6 @@
     "lib/jobs/reminder/medication-reminder-check.ts :: handleReminderCheck > withBackgroundEvent(arg)": 611,
     "lib/jobs/workout-insight-generate.ts :: runWorkoutInsightGenerate": 324,
     "lib/labs/biomarker-merge.ts :: runMerge": 254,
-    "lib/labs/ocr-extract.ts :: runOcrExtraction": 207,
-    "lib/labs/ocr-run.ts :: executeOcrExtraction": 206,
     "lib/measurements/consolidate-daily-mean.ts :: consolidateDailyMean": 278,
     "lib/measurements/consolidate-legacy-steps.ts :: consolidateLegacySteps": 263,
     "lib/measurements/dense-intraday-retention.ts :: runDenseIntradayRetention": 537,

src/__tests__/transcription-read-core-guard.test.ts

@@ -0,0 +1,133 @@
+/**
+ * Every report transcription reads through one core, under one set of limits.
+ *
+ * Two surfaces turn a report into rows a person reviews: the lab scan (Labs →
+ * Add → Scan a document) and the document extraction (Documents → a document
+ * → Extract lab values, and the lab staging after a document is read). Each
+ * used to carry its own copy of the read. When a long multi-page report
+ * outgrew one call, the lab scan learned to read it a page or a part at a
+ * time under a raised ceiling and to keep a cut answer's complete rows; the
+ * document extraction kept its single whole-document call, so the same report
+ * was read one way and failed the other as "Couldn't read the document. Try a
+ * clearer copy." Each surface passed its own tests.
+ *
+ * So the class is pinned here, structurally and by behaviour:
+ *   - a transcription read's output ceiling is set in exactly one file, the
+ *     core (`lib/documents/ocr-read.ts`), and its timeout is read only there
+ *     and in the deadline that plans the reads;
+ *   - both surfaces read through `runOcrReads` and dial no provider
+ *     themselves;
+ *   - both background runs plan their reads on the same deadline;
+ *   - a lab read and a document read, text and vision, go out with the same
+ *     limits.
+ *
+ * Its limit, stated: the structural cases are text matchers over `src/lib`. A
+ * third surface that builds its own completion under a different budget key
+ * would not be named here; one that copies the `ocrExtract` ceiling would.
+ */
+import { describe, expect, it } from "vitest";
+
+import { AI_BUDGETS } from "@/lib/ai/ai-budgets";
+import type { CompletionParams } from "@/lib/ai/types";
+import { inboundChunkParams } from "@/lib/documents/extract";
+import { ocrReadParams, type OcrChunk } from "@/lib/documents/ocr-read";
+import { ocrChunkParams } from "@/lib/labs/ocr-extract";
+
+import {
+  readSource,
+  readSourceFolder,
+  requireFloor,
+  stripComments,
+} from "./helpers/source-files";
+
+const LIB = new Map(
+  readSourceFolder("src/lib", {
+    floor: 1500,
+    include: (rel) => !rel.includes("__tests__") && !rel.endsWith(".test.ts"),
+  }).map(([rel, text]) => [rel, stripComments(text)] as const),
+);
+
+/** Files under `src/lib` whose code matches `pattern`, sorted. */
+function filesMatching(pattern: RegExp): string[] {
+  return [...LIB].filter(([, text]) => pattern.test(text)).map(([rel]) => rel);
+}
+
+/** The text of one top-level function in `source`, up to the next one. */
+function functionBody(source: string, name: string): string {
+  const start = source.search(new RegExp(`\\bfunction\\s+${name}\\s*\\(`));
+  expect(start, `function ${name}`).toBeGreaterThanOrEqual(0);
+  const rest = source.slice(start + 1);
+  const next = rest.search(/\n(?:export\s+)?(?:async\s+)?function\s+\w+\s*\(/);
+  return next < 0 ? rest : rest.slice(0, next);
+}
+
+describe("transcription read core", () => {
+  it("sets a read's output ceiling in the core and nowhere else", () => {
+    const setters = filesMatching(
+      /\bmaxTokens\s*:\s*AI_BUDGETS\.ocrExtract\.maxTokens\b/,
+    );
+    requireFloor("files setting the transcription ceiling", setters, 1);
+    expect(setters).toEqual(["documents/ocr-read.ts"]);
+  });
+
+  it("reads the per-read timeout in the core and in the deadline that plans the reads, nowhere else", () => {
+    const readers = filesMatching(/\bAI_BUDGETS\.ocrExtract\.timeoutMs\b/);
+    requireFloor("files reading the transcription timeout", readers, 2);
+    expect(readers).toEqual([
+      "documents/ocr-read.ts",
+      "jobs/document-ai-run.ts",
+    ]);
+  });
+
+  it("has both surfaces read through runOcrReads and dial no provider themselves", () => {
+    for (const rel of ["labs/ocr-extract.ts", "documents/extract.ts"]) {
+      const text = LIB.get(rel);
+      expect(text, rel).toBeDefined();
+      expect(text, rel).toMatch(/\brunOcrReads\s*\(/);
+      expect(text, rel).not.toMatch(/\.generateCompletion\s*\(/);
+    }
+  });
+
+  it("plans the lab scan run and the document extract run on the same deadline", () => {
+    const worker = stripComments(
+      readSource("src/lib/jobs/document-ai-run.ts", [
+        "function runLabsOcr",
+        "function runDocumentExtractRead",
+        "function ocrReadSchedule",
+      ]),
+    );
+    for (const fn of ["runLabsOcr", "runDocumentExtractRead"]) {
+      expect(functionBody(worker, fn), fn).toMatch(/\bocrReadSchedule\s*\(/);
+    }
+  });
+
+  it("sends a lab read and a document read, text and vision, under the same limits", () => {
+    const limits = (p: CompletionParams) => ({
+      maxTokens: p.maxTokens,
+      timeoutMs: p.timeoutMs,
+      temperature: p.temperature,
+      responseFormat: p.responseFormat,
+      reasoning: p.reasoning,
+    });
+    const expected = limits(ocrReadParams({ system: "s", user: "u" }));
+    expect(expected.maxTokens).toBe(AI_BUDGETS.ocrExtract.maxTokens);
+    expect(expected.timeoutMs).toBe(AI_BUDGETS.ocrExtract.timeoutMs);
+
+    const text: OcrChunk = { text: "Hb 14,2 g/dl", pages: [] };
+    const page: OcrChunk = {
+      images: [{ mediaType: "image/jpeg", dataBase64: "AAAA" }],
+      documents: [],
+      pages: [2],
+      pageTexts: [{ page: 2, text: "Hb 14,2 g/dl" }],
+    };
+    const reads = {
+      "lab text": ocrChunkParams("text", text, 0, 2),
+      "lab page": ocrChunkParams("vision", page, 1, 3, 3),
+      "document text": inboundChunkParams("text", text, 0, 2),
+      "document page": inboundChunkParams("vision", page, 1, 3, 3),
+    };
+    for (const [name, params] of Object.entries(reads)) {
+      expect(limits(params), name).toEqual(expected);
+    }
+  });
+});