docs(openapi): add dose-history import and its job-status poll

Documents POST /api/medications/intake/dose-history-import and
GET /api/medications/{id}/intake/import/{jobId}/status, including the
full 16-value MedicationImportSkipReason set the v1.33.0 shape widened
to, read straight from MEDICATION_IMPORT_SKIP_REASONS so the two never
drift again.

Also force-registers MedicationDoseHistoryImportFatalReason into
components.schemas: the 422 description names it, but nothing else
$refs it, and zod-openapi drops unreferenced schemas silently.

src/lib/openapi/routes/index.ts

@@ -52,7 +52,11 @@ import { ocrPaths } from "./ocr";
 import { measurementPaths } from "./measurements";
 import { measurementReminderPaths } from "./measurement-reminders";
 import { mentalHealthPaths } from "./mental-health";
-import { medicationPaths, medicationResource } from "./medications";
+import {
+  medicationPaths,
+  medicationResource,
+  medicationDoseHistoryImportFatalReasonEnum,
+} from "./medications";
 import { metaPaths } from "./meta";
 import { moodPaths } from "./mood";
 import { nutrientPaths } from "./nutrients";
@@ -121,8 +125,17 @@ export const openApiComponents: NonNullable<ZodOpenApiObject["components"]> = {
   // registration the standalone `Medication` component would never
   // emit. The iOS codegen reads from `Medication` directly for the
   // shared Swift struct backing both variants.
+  //
+  // `MedicationDoseHistoryImportFatalReason` lives here for the same reason:
+  // the dose-history-import 422 description names it as the closed set
+  // behind `errorCode`, but `ErrorEnvelope.meta.errorCode` is a bare string
+  // shared by every route, so nothing ever `$ref`s the enum itself. Without
+  // this forced registration the schema the description points to would be
+  // silently absent from the published spec.
   schemas: {
     Medication: medicationResource,
+    MedicationDoseHistoryImportFatalReason:
+      medicationDoseHistoryImportFatalReasonEnum,
   },
   securitySchemes: {
     bearerAuth: {

src/lib/openapi/routes/medications/index.ts

@@ -3,5 +3,8 @@
  *
  * Part of the OpenAPI route table; aggregated in `../index.ts`.
  */
-export { medicationResource } from "./schemas";
+export {
+  medicationResource,
+  medicationDoseHistoryImportFatalReasonEnum,
+} from "./schemas";
 export { medicationPaths } from "./paths";

src/lib/openapi/routes/medications/paths.ts

@@ -56,6 +56,8 @@ import {
   medicationListLayoutResult,
   doseHistoryQuery,
   doseHistoryResponse,
+  medicationIntakeImportJobStatusResponse,
+  medicationDoseHistoryImportResponse,
 } from "./schemas";
 
 // ── Bulk intake backfill (iOS SyncMode) — mirrors the route's
@@ -922,4 +924,105 @@ export const medicationPaths: NonNullable<ZodOpenApiObject["paths"]> = {
       },
     },
   },
+  "/api/medications/intake/dose-history-import": {
+    post: {
+      tags: ["Medications"],
+      summary: "Import a dose history exported from another medication tracker",
+      description:
+        "Reads a ten-column CSV export covering a whole regimen (multiple medications in one file), matching each row against the caller's existing medications by name — a medication is NEVER created from an import file. `text/csv` carries the actual dose history (`Date`/`Scheduled Date` in separate columns, mapped onto `takenAt`/`scheduledFor`). `application/json` is accepted at the content-type check but the documented JSON export shape always refuses at 422 (`json_carries_no_intake_time`): it has no field for when a dose was actually taken, and inferring one from the scheduled time would manufacture an on-time history the file never claimed. Body capped at 16 MB and 30 000 importable rows. `?dryRun=1` parses, matches, and returns the file verdict WITHOUT writing. A real submission queues the write onto the same background job the per-medication import uses (chunked, heartbeated, resumable) and returns 202 with a `statusUrl` to poll. Shares the 5/hour `import:` rate bucket with every other bulk import surface. Auth via cookie or Bearer.",
+      parameters: [
+        {
+          name: "dryRun",
+          in: "query",
+          required: false,
+          schema: { type: "string", enum: ["1", "true"] },
+          description:
+            "When `1` / `true`, parse + match + return the file verdict without queuing a write.",
+        },
+      ],
+      requestBody: {
+        required: true,
+        content: {
+          "text/csv": { schema: { type: "string" } },
+          "application/json": { schema: { type: "string" } },
+        },
+      },
+      responses: {
+        "200": {
+          description: "Dry-run verdict (nothing written).",
+          content: {
+            "application/json": {
+              schema: dataEnvelope(
+                medicationDoseHistoryImportResponse,
+                "MedicationDoseHistoryImportDryRunEnvelope",
+              ),
+            },
+          },
+        },
+        "202": {
+          description: "Import queued.",
+          content: {
+            "application/json": {
+              schema: dataEnvelope(
+                medicationDoseHistoryImportResponse,
+                "MedicationDoseHistoryImportQueuedEnvelope",
+              ),
+            },
+          },
+        },
+        "409": {
+          description: "A dose-history import is already in progress.",
+          content: { "application/json": { schema: errorEnvelope } },
+        },
+        "413": {
+          description: "File exceeds the 16 MB limit.",
+          content: { "application/json": { schema: errorEnvelope } },
+        },
+        "415": {
+          description: "Content-Type must be text/csv or application/json.",
+          content: { "application/json": { schema: errorEnvelope } },
+        },
+        "503": {
+          description: "Background worker is not available.",
+          content: { "application/json": { schema: errorEnvelope } },
+        },
+        ...stdResponses,
+        "422": {
+          description:
+            "The file could not be read (see `MedicationDoseHistoryImportFatalReason` for the closed `errorCode` set), or it holds more than 30 000 importable rows (`errorCode: too_many_rows`).",
+          content: { "application/json": { schema: errorEnvelope } },
+        },
+      },
+    },
+  },
+  "/api/medications/{id}/intake/import/{jobId}/status": {
+    get: {
+      tags: ["Medications"],
+      summary: "Poll a per-medication intake-import job",
+      description:
+        "Returns the queued/running/finished state of a per-medication intake-import job, scoped to the caller's own medication. A job created by the account-wide dose-history import (which carries no `medicationId`) never resolves through this path — poll its own `statusUrl` instead. v1.33.0 widened `result.skipReasons[].reason` / `progress.skipDetails[].reason` to the full 16-value `MedicationImportSkipReason` set. A foreign or unknown job/medication id is sealed as 404. Auth via cookie or Bearer.",
+      requestParams: {
+        path: z.object({ id: z.string(), jobId: z.string() }),
+      },
+      responses: {
+        "200": {
+          description: "Current job status.",
+          content: {
+            "application/json": {
+              schema: dataEnvelope(
+                medicationIntakeImportJobStatusResponse,
+                "MedicationIntakeImportJobStatusEnvelope",
+              ),
+            },
+          },
+        },
+        "404": {
+          description:
+            "No such job for this medication (or owned by another user).",
+          content: { "application/json": { schema: errorEnvelope } },
+        },
+        ...stdResponses,
+      },
+    },
+  },
 };

src/lib/openapi/routes/medications/schemas.ts

@@ -18,6 +18,7 @@ import {
   MEDICATION_ORDER_ID_MAX_LENGTH,
   MEDICATION_ORDER_MAX_ENTRIES,
 } from "@/lib/medication-list-layout";
+import { MEDICATION_IMPORT_SKIP_REASONS } from "@/lib/jobs/medication-intake-import";
 
 // ── Medications (v1.5 scheduling) ────────────────────────────────────
 //
@@ -929,3 +930,268 @@ export const medicationListLayoutResult = medicationListLayoutSchema
     description:
       "Resolved medications list presentation plus the optimistic-concurrency `updatedAt` token.",
   });
+
+// ── Intake-import jobs (v1.33.0) — the per-medication `/api/medications/{id}/
+// intake/import` route and the account-wide `/api/medications/intake/
+// dose-history-import` route both queue onto `MedicationIntakeImportJob` and
+// are polled through the SAME status projection, so the shapes below are
+// shared rather than duplicated per route.
+
+// The 16-reason closed set a row can be refused for, exactly as declared in
+// `MEDICATION_IMPORT_SKIP_REASONS`. `duplicate_in_file` / `already_recorded`
+// are dedup outcomes; `medication_not_found` / `medication_ambiguous` /
+// `medication_is_mirrored` / `missing_medication` are medication-matching
+// outcomes; `status_*` are source rows whose own status column said the dose
+// was never taken (a reminder, an unsent notification, an unrecognised
+// status); `missing_timestamp` / `missing_timezone_offset` /
+// `unreadable_timestamp` / `implausible_timestamp` are timestamp problems;
+// `unreadable_dosage` / `unreadable_row` are malformed cells.
+export const medicationImportSkipReasonEnum = z
+  .enum(MEDICATION_IMPORT_SKIP_REASONS)
+  .meta({
+    id: "MedicationImportSkipReason",
+    description:
+      "Why one source row did not become an intake row. Closed 16-value set: duplicate_in_file, already_recorded, medication_not_found, medication_ambiguous, medication_is_mirrored, status_no_dose_information, status_reminder_event, status_notification_not_sent, status_unknown, missing_timestamp, missing_timezone_offset, unreadable_timestamp, implausible_timestamp, missing_medication, unreadable_dosage, unreadable_row.",
+  });
+
+const medicationImportSkipGroup = z
+  .object({
+    reason: medicationImportSkipReasonEnum,
+    count: z.number().int().min(1),
+  })
+  .meta({ id: "MedicationImportSkipGroup" });
+
+const medicationImportSkipDetail = z
+  .object({
+    line: z.number().int().describe("1-based source-file row ordinal."),
+    reason: medicationImportSkipReasonEnum,
+  })
+  .meta({
+    id: "MedicationImportSkipDetail",
+    description:
+      "One refused source row, reduced to the ordinal + reason — never the row's own content (no medication name, timestamp, or dosage text).",
+  });
+
+// The terminal, persisted result. Reconstructed through a strict allowlist
+// (`projectMedicationImportResult`) before it ever reaches a polling
+// response, so a compromised or legacy JSON blob cannot smuggle anything
+// past this shape.
+export const medicationImportResult = z
+  .object({
+    imported: z.number().int().min(0),
+    skipped: z
+      .number()
+      .int()
+      .min(0)
+      .describe("Total entries that did not become a row."),
+    skipReasons: z
+      .array(medicationImportSkipGroup)
+      .describe(
+        "One entry per reason that actually occurred, count descending then reason — never the same sentence repeated per row. Empty when nothing was skipped.",
+      ),
+    skipDetails: z
+      .array(medicationImportSkipDetail)
+      .optional()
+      .describe(
+        "First bounded source-row details (capped at 200). Absent on a legacy job run before this field existed.",
+      ),
+    skippedDetailsOmitted: z
+      .number()
+      .int()
+      .min(0)
+      .optional()
+      .describe(
+        "Count of additional skipped rows beyond the `skipDetails` cap, counted without retaining their contents. Absent when nothing was omitted.",
+      ),
+  })
+  .meta({
+    id: "MedicationImportResult",
+    description:
+      "The finished job's outcome. Null while the job is still queued or running — `status` + `progress` are the fields to read meanwhile.",
+  });
+
+// The in-flight progress blob a running/queued job carries. Written whole at
+// admission time and advanced by the worker in place; unlike `result` it is
+// NOT re-validated on read (stored as a bare JSON column), so this documents
+// the shape the job worker actually writes rather than a server-enforced
+// contract.
+export const medicationImportProgress = z
+  .object({
+    processed: z
+      .number()
+      .int()
+      .min(0)
+      .describe("Entries the worker has walked so far, in file order."),
+    total: z.number().int().min(0).describe("Entries queued for this job."),
+    imported: z
+      .number()
+      .int()
+      .min(0)
+      .describe("Entries that have reached the database so far."),
+    skippedByReason: z
+      .partialRecord(medicationImportSkipReasonEnum, z.number().int().min(1))
+      .describe(
+        "Running skip tally by reason. A reason with no skips yet is absent from the object, never present as 0.",
+      ),
+    skipDetails: z
+      .array(medicationImportSkipDetail)
+      .optional()
+      .describe("Same bounded detail list as the finished `result`."),
+    skippedDetailsOmitted: z.number().int().min(0).optional(),
+    touchedDays: z
+      .array(z.string())
+      .describe(
+        "Local calendar days (YYYY-MM-DD) touched so far, driving the chunked rollup recompute behind the write.",
+      ),
+    rollupProcessed: z
+      .number()
+      .int()
+      .min(0)
+      .describe("Of `touchedDays`, how many have had their rollup recomputed."),
+  })
+  .meta({
+    id: "MedicationImportProgress",
+    description:
+      "In-flight job progress, written whole at admission and advanced in place by the worker. Present from the moment the job is created (queued), so `processed: 0, total: N` is the shape a client sees immediately after kickoff, not an empty object.",
+  });
+
+export const medicationIntakeImportJobStatusResponse = z
+  .object({
+    jobId: z.string(),
+    status: z
+      .enum(["queued", "running", "done", "failed"])
+      .describe(
+        "queued — admitted, not yet picked up. running — the worker is walking entries. done — finished (see `result` for the outcome, even a fully-refused run). failed — the worker could not complete the run (see `failureReason`).",
+      ),
+    progress: medicationImportProgress,
+    result: medicationImportResult
+      .nullable()
+      .describe("Set only once `status` is `done`; null before that."),
+    failureReason: z
+      .string()
+      .nullable()
+      .describe("Set only when `status` is `failed`; null otherwise."),
+    createdAt: z.iso.datetime({ offset: true }),
+    startedAt: z.iso
+      .datetime({ offset: true })
+      .nullable()
+      .describe("When the worker picked the job up; null while still queued."),
+    completedAt: z.iso
+      .datetime({ offset: true })
+      .nullable()
+      .describe("When the job reached `done` or `failed`; null while active."),
+  })
+  .meta({
+    id: "MedicationIntakeImportJobStatusResponse",
+    description:
+      "Poll target for a queued intake-import job — shared by the per-medication import and the account-wide dose-history import, so a client polls both through one shape.",
+  });
+
+// ── Account-wide dose-history import (v1.33.0) ───────────────────────
+
+// The file-level verdict, identical for a dry run and a real submission.
+const medicationDoseHistoryImportFile = z
+  .object({
+    rowsRead: z
+      .number()
+      .int()
+      .min(0)
+      .describe("Data rows read from the file, refusals included."),
+    queued: z
+      .number()
+      .int()
+      .min(0)
+      .describe("Rows that parsed and matched a medication."),
+    refused: z
+      .number()
+      .int()
+      .min(0)
+      .describe("Total rows that did not queue, across every reason."),
+    refusedByReason: z
+      .array(medicationImportSkipGroup)
+      .describe(
+        "One entry per reason that actually occurred, count descending then reason.",
+      ),
+    unmatchedMedications: z
+      .array(z.string())
+      .max(50)
+      .describe(
+        "Medication names the file used that match nothing on the record, deduplicated, first-seen order, capped at 50.",
+      ),
+    ambiguousMedications: z
+      .array(z.string())
+      .max(50)
+      .describe(
+        "Names that match more than one medication, so no single one can be meant. Capped at 50.",
+      ),
+    mirroredMedications: z
+      .array(z.string())
+      .max(50)
+      .describe(
+        "Names that match only a medication mirrored from an external list. Capped at 50.",
+      ),
+    unknownColumns: z
+      .array(z.string())
+      .max(50)
+      .describe("Header cells the importer has no verdict for. Capped at 50."),
+    codingsNotRead: z
+      .number()
+      .int()
+      .min(0)
+      .describe("Rows whose `Codings` cell held something left unread."),
+    fromArchivedMedications: z
+      .number()
+      .int()
+      .min(0)
+      .describe(
+        "Rows the file marked as belonging to a medication archived in the source.",
+      ),
+  })
+  .meta({ id: "MedicationDoseHistoryImportFile" });
+
+export const medicationDoseHistoryImportResponse = z
+  .object({
+    dryRun: z.boolean(),
+    jobId: z
+      .string()
+      .nullable()
+      .describe("Null on a dry run (nothing was queued)."),
+    status: z
+      .literal("queued")
+      .optional()
+      .describe("Present only on a real (non-dry-run) submission."),
+    statusUrl: z
+      .string()
+      .nullable()
+      .describe(
+        "Relative poll path (`/api/medications/intake/dose-history-import/{jobId}/status`). Null on a dry run.",
+      ),
+    file: medicationDoseHistoryImportFile,
+  })
+  .meta({
+    id: "MedicationDoseHistoryImportResponse",
+    description:
+      "The file-level verdict, plus the queued job pointer on a real submission. A dry run (`?dryRun=1`) returns the same `file` verdict with `jobId`/`statusUrl` null and no `status` field — parses, matches, and reports without writing anything.",
+  });
+
+// The 5 file-level (whole-submission) refusal reasons — distinct from the
+// 16 per-row `medicationImportSkipReasonEnum` above. `application/json` is
+// accepted at the content-type check, but the documented JSON export shape
+// always ends in `json_carries_no_intake_time`: it carries `scheduledDate` +
+// `status` + `dosage` but no field for when a dose was actually taken, and
+// writing the scheduled time as `takenAt` would manufacture an on-time
+// history the file never claimed. `text/csv` is the only shape that
+// actually imports.
+export const medicationDoseHistoryImportFatalReasonEnum = z
+  .enum([
+    "empty_file",
+    "missing_required_columns",
+    "unreadable_json",
+    "json_not_an_array",
+    "json_carries_no_intake_time",
+  ])
+  .meta({
+    id: "MedicationDoseHistoryImportFatalReason",
+    description:
+      "Whole-file refusal reason on the 422 `errorCode`. empty_file — no data rows. missing_required_columns — the CSV header is missing a column the parser needs. unreadable_json — the body did not parse as JSON. json_not_an_array — parsed JSON is not an array (and no nested array was found under any key). json_carries_no_intake_time — well-formed JSON export input, refused by design (see above).",
+  });