docs: trim 1.2.0:6 Current state to a lean handoff summary

Add a single-session AI flow alongside program generation: describe a
workout in plain words and get a ready-to-log workout back — exercises
with suggested weights, target reps, and set counts grounded in the
user's recent history. The suggestion can be inline-edited or refined
by sending a follow-up instruction back to the model, then "Use this
workout" pre-fills the normal New Workout form (nothing persists until
the user saves through the regular path).

Why reuse, not fork: the existing program-generation spine (detached
background runner, SSE streaming, lenient-JSON preview, 5 providers,
history context, library name->id mapping) already does the hard parts.
A new AIGeneration.kind discriminant ("program" | "workout", default
"program" via boot-time guarded ALTER) selects the parser and keeps the
ephemeral workout rows out of the program-shaped AI history. Refine is a
fresh generation seeded with the prior suggestion (validated through the
same schema before it re-enters the prompt).

Hand-off is sessionStorage -> /main/workouts/new?from=ai -> AiWorkoutPrefill,
which expands each suggestion into N sets and maps effort by cardio-ness
(Gear for cardio, RPE for strength). EditWorkoutData.id is now optional so
the prefill CREATEs rather than PATCHing a nonexistent id. The AI suggests
each weight in that exercise's effective logging unit (the library JSON
carries a per-exercise unit) so the stored number and unit never diverge.

Built + sideloaded to immense-voyage.local as 1.2.0:6; on-box ALTER and
non-root launch confirmed via start-cli. tsc clean (app + packaging),
251 tests pass, next build + s9pk build succeed.

AGENTS.md

@@ -63,6 +63,11 @@ Build/sideload the s9pk (from `start9/0.4/`): `make x86` then `make install`. Ta
 
 Both `install` and `publish` read host/registry config from `~/.startos/config.yaml`, which is **not in the repo** — verify against the live setup, not from a checkout.
 
+**Verify on-box state read-only via `start-cli`** (the same host config) instead of punting to the StartOS web UI — used this way to confirm the 1.2.0:4/:5 ALTERs and a persisted set:
+- `start-cli package installed-version proof-of-work` — what version the box actually runs.
+- `start-cli package logs proof-of-work --limit N | grep -iE "adding missing column|as nextjs|error"` — confirms boot ALTERs ran (each logs once) and the non-root launch.
+- `start-cli package attach proof-of-work -- sqlite3 /data/app.db "<SELECT …>"` — read-only query of the live app DB (e.g. confirm a new column exists / a value persisted). **SELECT only**; never mutate prod data this way.
+
 Canonical publish path for this project: `~/.proof-of-work/publish.sh` (builds, uploads to FileBrowser, registers) — separate from the generic `make publish`. Unpublish: `~/.proof-of-work/unpublish.sh`.
 
 `npm run db:seed` (= `tsx prisma/seed.ts`) seeds **only the `InstanceSettings` singleton** — deliberately NO users and NO curated library (the library attaches at admin-creation and via the boot-time ensure). It is **live, not dead** — invoked at Docker image-build time (`start9/0.4/Dockerfile`) and the local-dev first-run path. `npm run create-admin` (= `tsx scripts/create-admin.ts`) is the local-dev equivalent of the StartOS "Set admin credentials" action: creates the first admin + seeds their library; `--force` to reset/promote an existing account. Runtime first-boot/upgrade seeding is handled separately by `docker_entrypoint.sh`.
@@ -72,6 +77,8 @@ Canonical publish path for this project: `~/.proof-of-work/publish.sh` (builds,
 - **Versioning is ExVer**: `1.1.0:4` (note the colon). Every release = a new `start9/0.4/startos/versions/vMAJOR.MINOR.PATCH.N.ts` file, imported into `versions/index.ts` and promoted to `current` (previous `current` moves into `other[]`).
 - **Bump the version BEFORE building the s9pk** — Start9 0.4 won't recognize a rebuild as an update otherwise.
 - **Schema changes are additive ALTERs in `docker_entrypoint.sh`**, guarded by `PRAGMA table_info` checks. Keep `schema.prisma` in sync as the mirror, but the entrypoint is what migrates live `/data`. Never write a destructive migration.
+- **Adding a first-class numeric set metric** (precedent: `watts`, 1.2.0:4): mirror `calories` end-to-end — `schema.prisma` column + `prisma generate`; guarded additive `ALTER` in `docker_entrypoint.sh`; zod field + insert in all **5 set-write paths** (`workouts` POST, `workouts/[id]` PATCH, `workouts/[id]/sets`, `workouts/import/save`, `me/import`); `SetRow.tsx` (`show*`/state/`emitUpdate`/`buildSummary`/`firstField`/input) + `WorkoutForm.tsx` (set interfaces, `buildPayload`, `handleUpdateSet`, `initial*` prop, has-data checks); read summary in `app/main/workouts/[id]/page.tsx` + edit mapping in `app/main/workouts/new/page.tsx`; CSV export/parse + `page-csv` payload; field-option label lists (`lib/exerciseOptions.ts`, `app/main/exercises/[id]/page.tsx`, `ExercisePicker.tsx`). The `inputFields` token == the column name; the human label lives in those option lists (token `watts` → "Avg. watts"). `me/export` rides the 1:1 Prisma dump automatically. Add a round-trip test in `tests/routes-crud.test.ts`.
+- **Logged-set effort is Gear or RPE, by cardio-ness** (1.2.0:5): the effort select is always shown (not an `inputFields` token). Cardio exercises log breathing **Gear** (`SetLog.gear`, 1–5); everything else logs **RPE** (`SetLog.rpe`, 6–10). The switch is `isCardioExercise(exercise)` (`lib/exerciseOptions.ts`): `type === "cardio"` OR `muscleGroups` contains "cardio". `SetRow` takes an `isCardio` prop (from `WorkoutForm`) and renders one; both are always emitted (the hidden one stays empty). Distinct from program/AI **target**-RPE (`ProgramExercise.rpe`), which is unrelated and unaffected.
 - **Commit subject** = `vX.Y.Z:N — short summary`, imperative, body explains the *why*.
 - **Git remote is self-hosted** (a private Start9 registry + a FileBrowser artifact host), NOT GitHub. The actual registry/file-host URLs are constants in `~/.proof-of-work/{publish,unpublish}.sh`; FileBrowser creds live in `~/.keysat/filebrowser.env` (outside the repo, gitignored). Default branch is `master`.
 - **Authorization tiers**: whole-instance routes (`settings/{export,import}-db`) are **admin-only** (`!user.isAdmin → 403`); per-user data routes scope by `user.id`. Custom-URL AI providers (Ollama, OpenAI-compatible — anything with `requiresBaseUrl`) are **admin-only** (SSRF surface); fixed-URL cloud providers (claude/openai/gemini) stay per-user. Gate the server route AND hide the control in the UI.
@@ -105,13 +112,13 @@ Canonical publish path for this project: `~/.proof-of-work/publish.sh` (builds,
 
 ## Current state
 
-Latest version is **1.2.0:4** — **avg. watts is now a first-class set field**: promotes average watts (assault bike / rower / ski erg) from a free-text `customMetrics` JSON entry to a real nullable `SetLog.watts` column, written through every set path (create / PATCH / add-sets / import-save / account-import) and shown everywhere as "Avg. watts" with a numeric input. This is the **first schema change in the 1.2.0 line**: the column is added by the boot-time additive `ALTER` in `docker_entrypoint.sh` (guarded, idempotent); legacy watts in `customMetrics` stays readable and migrates to the column on next save (no data migration). Internal token/column is `watts` (matches the existing assault-bike `inputFields` token + legacy `customMetrics` key → zero exercise migration); only ever *displayed* as "Avg. watts". **Built + sideloaded** to the StartOS box (`immense-voyage.local`, 2026-06-16, on `master`) as `proof-of-work_x86_64.s9pk` (80M, git `390aaf5`). Verified before build: tsc clean (app + packaging), lint clean (only pre-existing warnings), **223 tests pass**, `next build` succeeds. Registry empty, **publishing parked** (sideload-only via `make install`).
+Latest version is **1.2.0:6** — **AI "generate today's workout"**: describe one session in plain words → a streamed, ready-to-log workout (suggested weights/reps/set-counts grounded in 90-day history) → inline-edit or **Refine** (round-trips changes back to the LLM) → **Use this workout** pre-fills the New Workout form (nothing persists until you save). Reuses the program-generation spine via a new `AIGeneration.kind` discriminant (`"program" | "workout"`); workout rows are ephemeral (the saved Workout is the durable record) so they're filtered out of the program-shaped AI History. New `AIGeneration.kind` column (default "program") via boot-time guarded `ALTER`. **Architecture + hand-off (sessionStorage `?from=ai` → `AiWorkoutPrefill`, `EditWorkoutData.id?`→CREATE, per-exercise weight-unit grounding) are documented in `docs/guides/ai-subsystem.md` → "Two generation kinds".** **Built + sideloaded** (`immense-voyage.local`, 2026-06-19, `master`) as `proof-of-work_x86_64.s9pk` (80M). tsc clean (app + packaging), lint clean (pre-existing only), **251 tests pass**, `next build` + s9pk build succeed. Registry empty, **publishing parked** (sideload-only via `make install`).
 
-Recent prior ships (1.2.0 line): **1.2.0:3** — P3 hardening (login timing oracle + `exerciseId` ownership). **1.2.0:2** — iOS Safari login/signup first-tap retry (`lib/retryAction.ts`). **1.2.0:1** — Next 15 / React 19 upgrade. (1.2.0:1–3 had no schema/data change.)
+**Confirmed on-box (2026-06-19, via `start-cli`):** box runs `1.2.0:6`; entrypoint logged `adding AIGeneration.kind (default 'program')` once, then launched `as nextjs` with no errors (clears the long-standing non-root check); read-only `SELECT` confirms the `AIGeneration.kind` column exists and the existing generation row backfilled to `program`. Recent prior ships (1.2.0 line): **1.2.0:5** Gear replaces RPE for cardio; **1.2.0:4** watts as first-class set field; **1.2.0:3** P3 hardening (login timing oracle + `exerciseId` ownership).
 
-**Pending on-box check:** confirm 1.2.0:4 boots clean in StartOS → Logs — on the existing-data upgrade path the entrypoint logs `adding missing column SetLog.watts` (one-shot; no-op thereafter), no permission errors — and that logging an Assault Bike set with Avg. watts saves and displays as `157 W`. Still also pending from earlier ships: the 1.2.0:3 non-root boot (entrypoint logs `launching … as nextjs`, app writes `/data` as uid 1001 — also clears the long-standing 1.1.0:9 non-root check), **and** the 1.2.0:2 Safari first-tap proof (log in from Safari on iPhone/iPad; first Sign In tap works — `-1005` in Web Inspector confirms the retry path, anything else points at a proxy↔container keep-alive mismatch).
+**No on-box checks pending.** Known bug (tracked in `ROADMAP.md` → Known bugs): the **1.2.0:2** Safari first-tap retry did NOT fix the mobile-Safari first-login failure — reproduced through 1.2.0:5 (the workout feature didn't touch auth), first tap shows "An unexpected error occurred", second tap works. Diagnosis captured; the fix is gated on one data point — the first failed request's error code from Safari Web Inspector (`-1005` → client delayed-retry; `502`/`503` → Node keep-alive tuning).
 
-Working: workout logging, programs (manual + AI), multi-user, curated library, full AI subsystem (5 providers, multi-config, background generation, history detail, cost/duration, Ollama auto-detect, infinite-scroll exercise history).
+Working: workout logging, programs (manual + AI), multi-user, curated library, full AI subsystem (5 providers, multi-config, background generation, **single-workout generation + refine**, history detail, cost/duration, Ollama auto-detect, infinite-scroll exercise history).
 
 Next steps (priority order):
 1. **Finish the P3 hardening batch** (`ROADMAP.md` → Security & hardening — timing oracle + exerciseId ownership now DONE): CSP `unsafe-eval`, `/api/health` info disclosure, rate-limit map leak, configurable/shorter sessions (currently 30-day), text max-length. Also unify the 3rd JSON-parse pattern (`try{json}catch{→{}}`) in `programs/[id]/days/[dayId]/start`.

ROADMAP.md

@@ -2,6 +2,14 @@
 
 Longer-term backlog. Near-term state + next steps live in `AGENTS.md` → Current state.
 
+## Known bugs
+
+- **Mobile-Safari first-login-tap fails ("An unexpected error occurred"); second tap works.** Reproduced on iPhone/iPad Safari against 1.2.0:5 (desktop Safari untested — user declined). The first Sign In tap fails, a second manual tap succeeds. **1.2.0:2's `retryOnTransportError` does NOT fix it.** Diagnosis so far: `LoginForm` only surfaces that error when *both* the initial action call and its in-tap retry throw, so the immediate retry isn't escaping the bad connection — only a fresh user-initiated tap does. Box app logs show no server-side error/500/reset around the attempt, so it's a transport-layer failure, not an app bug.
+  - **Gating data (do this first):** capture the first failed request's error in Safari Web Inspector (iOS→Mac, Network/Console tab). The code picks the fix:
+    - `-1005` "The network connection was lost" → client-side stale keep-alive socket. Fix = a *delayed* retry (let Safari tear down the dead socket before retrying), not the current instant one.
+    - `502`/`503` → StartOS-proxy↔Node keep-alive mismatch (Node closing idle conns the proxy reuses). Fix = raise Node `keepAliveTimeout`/`headersTimeout` server-side; a client retry only masks it.
+  - Files: `lib/retryAction.ts`, `app/auth/login/LoginForm.tsx`, `app/auth/signup/SignupForm.tsx`.
+
 ## AI quality
 
 - Tiered prompt formatting (also the immediate next step): JSON-Schema output enforcement via Ollama `format` and OpenAI `response_format`; pipe-separated library rows; XML-tagged prompt sections; Ollama-only few-shot example; stable prefix first for prompt-cache hits.
@@ -25,6 +33,7 @@ Done in 1.1.0:9 (P2 batch): input-validation 500s → 400 (`lib/http.ts readJson
 
 - Adherence tracking: compare logged workouts against the planned `ProgramDay` (the `programDayId` link already exists).
 - Per-user export/import polish and scheduled backups.
+- CSV export↔import round-trip: export writes `setX`-prefixed headers (`setCalories`/`setWatts`/`setNotes`) the importer doesn't read (it expects `calories`/`watts`/`notes`), so the app's own CSV export silently drops those on re-import (calories long-standing; watts since 1.2.0:4). Fix by aligning export header names with the parser, or adding the prefixed names as `knownColumns` aliases. (JSON account export/import round-trips fine.)
 - Charts/progress views over history (the data and 1RM estimates already exist).
 
 ## Hygiene

docs/guides/ai-subsystem.md

@@ -20,6 +20,29 @@ generate/generations route handlers). Whole-repo rules live in `AGENTS.md`.
 - Multi-config: `AIConfigProfile` rows per user; `UserPreferences.activeAIConfigId`
   points at the active one and is mirrored into the legacy `ai*` columns for back-compat.
 
+## Two generation kinds (`AIGeneration.kind`)
+
+The runner spine is shared by two output shapes, discriminated by `AIGeneration.kind`
+("program" | "workout", default "program"). The runner picks the parser by kind and
+stores the JSON in the (reused) `parsedProgram` column.
+
+- **program** (`kind: 'program'`) — `generate/route.ts` → `programSchema.ts`
+  (`PROGRAM_OUTPUT_SHAPE` / `parseAIProgram`). Applied to DB rows via `apply.ts`.
+  Shown in AI · History (which filters `kind: 'program'`).
+- **workout** (`kind: 'workout'`) — `generate-workout/route.ts` (uses
+  `workoutPrompt.ts` + `workoutSchema.ts`: `WORKOUT_OUTPUT_SHAPE` / `parseAIWorkout`).
+  A single day's session. **No server-side apply**: the client (`GenerateWorkoutClient.tsx`)
+  stashes the reviewed suggestion in `sessionStorage` and routes to
+  `/main/workouts/new?from=ai`, where `AiWorkoutPrefill.tsx` expands it (via
+  `workoutDraft.ts::buildPrefillExercises`) and pre-fills the normal `WorkoutForm` —
+  nothing persists until the user saves through the regular workout path.
+  **Refine = a new workout generation** seeded with the prior suggestion JSON
+  (`priorWorkout` in the route body → REVISION mode in `workoutPrompt.ts`). These rows
+  are ephemeral, so they're excluded from the program-shaped AI · History.
+- Adding a new kind: extend the union in `KickoffOpts`, add a parser + output-shape,
+  branch the parser selection in `generationRunner.ts`, and decide whether it belongs in
+  History (filtered by kind).
+
 ## Provider abstraction
 
 - Each provider yields an async iterable of `GenerateChunk` (`text` / `usage` / `done` /

proof-of-work/app/api/ai/generate-workout/route.ts

@@ -0,0 +1,149 @@
+import { NextRequest, NextResponse } from 'next/server';
+import { z } from 'zod';
+import { getCurrentUser } from '@/lib/auth';
+import { prisma } from '@/lib/prisma';
+import { WORKOUT_OUTPUT_SHAPE, aiWorkoutSchema } from '@/lib/ai/workoutSchema';
+import {
+  buildHistorySummary,
+  formatHistoryContext,
+} from '@/lib/ai/historyContext';
+import { buildWorkoutSystemPrompt } from '@/lib/ai/workoutPrompt';
+import { kickoffGeneration } from '@/lib/ai/generationRunner';
+
+/**
+ * POST /api/ai/generate-workout
+ *
+ * Kicks off a background runner (kind="workout") that streams a single
+ * day's workout suggestion, and returns the generation id. The caller
+ * subscribes via GET /api/ai/generations/[id]/stream (SSE) — the same
+ * spine as program generation.
+ *
+ * Body:
+ *   { userInput: string, includeHistory?: boolean, priorWorkout?: AIWorkout }
+ *
+ * `priorWorkout` switches the prompt into REVISION mode: userInput is the
+ * change instruction and the model re-emits the full revised workout.
+ *
+ * Response:
+ *   201 { id: "...generationId..." }
+ *   400 { error: "..." }
+ */
+
+const bodySchema = z.object({
+  userInput: z.string().min(1),
+  includeHistory: z.boolean().optional().default(false),
+  // The current suggestion, when refining. Validated against the same
+  // shape the model emits so we only ever feed it well-formed JSON.
+  priorWorkout: aiWorkoutSchema.optional().nullable(),
+});
+
+export const dynamic = 'force-dynamic';
+
+export async function POST(request: NextRequest) {
+  const user = await getCurrentUser();
+  if (!user) {
+    return NextResponse.json({ error: 'Unauthorized' }, { status: 401 });
+  }
+
+  const body = await request.json().catch(() => ({}));
+  const parsed = bodySchema.safeParse(body);
+  if (!parsed.success) {
+    return NextResponse.json(
+      { error: 'Invalid body', details: parsed.error.errors },
+      { status: 400 },
+    );
+  }
+
+  const prefs = await prisma.userPreferences.findUnique({
+    where: { userId: user.id },
+  });
+  if (!prefs?.aiProvider || !prefs?.aiModel) {
+    return NextResponse.json(
+      {
+        error:
+          'AI is not configured. Open Settings → AI integration and pick a provider + model.',
+      },
+      { status: 400 },
+    );
+  }
+
+  const weightUnit = (prefs.defaultWeightUnit as 'lbs' | 'kg') || 'lbs';
+
+  // Library for the prompt. We include each exercise's effective logging
+  // unit (`defaultWeightUnit || "lbs"` — the exact unit the saved workout
+  // will store, see WorkoutForm.buildPayload) so the model suggests the
+  // weight NUMBER in that unit. Without this the model would assume the
+  // user's global preferred unit, which diverges for per-exercise unit
+  // overrides (e.g. kettlebells in kg) and silently mislabels the weight.
+  const exercises = await prisma.exercise.findMany({
+    where: { userId: user.id },
+    select: {
+      id: true,
+      name: true,
+      type: true,
+      muscleGroups: true,
+      defaultWeightUnit: true,
+    },
+  });
+  const libraryJson = JSON.stringify(
+    exercises.map((e) => ({
+      id: e.id,
+      name: e.name,
+      type: e.type,
+      unit: e.defaultWeightUnit || 'lbs',
+      muscleGroups: (() => {
+        try {
+          return JSON.parse(e.muscleGroups);
+        } catch {
+          return [];
+        }
+      })(),
+    })),
+  );
+
+  // History context if requested.
+  let historyBlock = '';
+  if (parsed.data.includeHistory) {
+    const summary = await buildHistorySummary(prisma, user.id);
+    historyBlock = formatHistoryContext(summary);
+  }
+
+  const isLocalModel = prefs.aiProvider === 'ollama';
+  const priorWorkoutJson = parsed.data.priorWorkout
+    ? JSON.stringify(parsed.data.priorWorkout)
+    : undefined;
+
+  const basePrompt = buildWorkoutSystemPrompt({
+    weightUnit,
+    hasHistoryContext: parsed.data.includeHistory,
+    isLocalModel,
+    priorWorkoutJson,
+  });
+
+  const systemPrompt = `${basePrompt}
+
+# OUTPUT SHAPE
+
+${WORKOUT_OUTPUT_SHAPE}
+
+# LIBRARY (use these exerciseIds; do not invent ids)
+
+${libraryJson}${historyBlock}`;
+
+  const id = await kickoffGeneration({
+    prisma,
+    userId: user.id,
+    kind: 'workout',
+    templateId: null,
+    templateName: priorWorkoutJson ? 'Workout (refine)' : 'Workout',
+    userInput: parsed.data.userInput,
+    systemPrompt,
+    userPrompt: parsed.data.userInput,
+    provider: prefs.aiProvider,
+    model: prefs.aiModel,
+    apiKey: prefs.aiApiKey,
+    baseUrl: prefs.aiBaseUrl,
+  });
+
+  return NextResponse.json({ id }, { status: 201 });
+}

proof-of-work/app/api/ai/generate/route.ts

@@ -147,6 +147,7 @@ ${libraryJson}${historyBlock}`;
   const id = await kickoffGeneration({
     prisma,
     userId: user.id,
+    kind: 'program',
     templateId: template?.id ?? null,
     templateName: template?.name ?? null,
     userInput: parsed.data.userInput,

proof-of-work/app/api/ai/generations/route.ts

@@ -16,7 +16,8 @@ export async function GET(request: NextRequest) {
   const offset = Math.max(parseInt(sp.get('offset') || '0'), 0);
 
   const rows = await prisma.aIGeneration.findMany({
-    where: { userId: user.id },
+    // Program history only; workout-kind rows are ephemeral (see history page).
+    where: { userId: user.id, kind: 'program' },
     orderBy: { createdAt: 'desc' },
     take: limit + 1,
     skip: offset,

proof-of-work/app/api/import/parse/route.ts

@@ -21,6 +21,7 @@ interface ParsedSet {
   calories?: number;
   watts?: number;
   rpe?: number;
+  gear?: number;
   customMetrics?: Record<string, string>;
   notes?: string;
 }
@@ -139,6 +140,7 @@ export async function POST(request: NextRequest) {
       "calories",
       "watts",
       "rpe",
+      "gear",
       "notes",
       "custom_metrics_json",
       "custommetricsjson",
@@ -203,6 +205,7 @@ export async function POST(request: NextRequest) {
         const calories = parseIntMaybe(row.calories);
         const watts = parseIntMaybe(row.watts);
         const rpe = parseIntMaybe(row.rpe);
+        const gear = parseIntMaybe(row.gear);
 
         const customMetrics: Record<string, string> = {};
         const customJson = row.custom_metrics_json || row.custommetricsjson;
@@ -258,6 +261,7 @@ export async function POST(request: NextRequest) {
           calories,
           watts,
           rpe,
+          gear,
           customMetrics: Object.keys(customMetrics).length > 0 ? customMetrics : undefined,
           notes: notes || undefined,
         });

proof-of-work/app/api/me/import/route.ts

@@ -51,6 +51,7 @@ const setLogImport = z.object({
   weight: z.number().nullable().optional(),
   weightUnit: z.string().optional(),
   rpe: z.number().int().nullable().optional(),
+  gear: z.number().int().nullable().optional(),
   durationSeconds: z.number().int().nullable().optional(),
   distance: z.number().nullable().optional(),
   distanceUnit: z.string().nullable().optional(),
@@ -202,6 +203,7 @@ export async function POST(request: NextRequest) {
               weight: s.weight ?? null,
               weightUnit: s.weightUnit ?? 'lbs',
               rpe: s.rpe ?? null,
+              gear: s.gear ?? null,
               durationSeconds: s.durationSeconds ?? null,
               distance: s.distance ?? null,
               distanceUnit: s.distanceUnit ?? null,

proof-of-work/app/api/settings/export-csv/route.ts

@@ -72,6 +72,7 @@ export async function GET() {
       "setCalories",
       "setWatts",
       "rpe",
+      "setGear",
       "setNotes",
       "customMetricsJson",
     ];
@@ -104,6 +105,7 @@ export async function GET() {
         set.calories ?? "",
         set.watts ?? "",
         set.rpe ?? "",
+        set.gear ?? "",
         set.notes ?? "",
         set.customMetrics ?? "",
       ];

proof-of-work/app/api/workouts/[id]/route.ts

@@ -57,6 +57,7 @@ const setSchema = z.object({
   weight: z.number().optional().nullable(),
   weightUnit: z.string().default("lbs"),
   rpe: z.number().int().min(1).max(10).optional().nullable(),
+  gear: z.number().int().min(1).max(5).optional().nullable(),
   durationSeconds: z.number().int().positive().optional().nullable(),
   distance: z.number().positive().optional().nullable(),
   distanceUnit: z.string().optional().nullable(),
@@ -153,6 +154,7 @@ export async function PATCH(
               weight: set.weight ?? undefined,
               weightUnit: set.weightUnit,
               rpe: set.rpe ?? undefined,
+              gear: set.gear ?? undefined,
               durationSeconds: set.durationSeconds ?? undefined,
               distance: set.distance ?? undefined,
               distanceUnit: set.distanceUnit ?? undefined,

proof-of-work/app/api/workouts/[id]/sets/route.ts

@@ -14,6 +14,7 @@ const addSetsSchema = z.object({
       weight: z.number().optional(),
       weightUnit: z.string().default("lbs"),
       rpe: z.number().int().min(1).max(10).optional(),
+      gear: z.number().int().min(1).max(5).optional(),
       durationSeconds: z.number().int().positive().optional(),
       distance: z.number().positive().optional(),
       distanceUnit: z.string().optional(),
@@ -80,6 +81,7 @@ export async function POST(
         weight: set.weight,
         weightUnit: set.weightUnit,
         rpe: set.rpe,
+        gear: set.gear,
         durationSeconds: set.durationSeconds,
         distance: set.distance,
         distanceUnit: set.distanceUnit,

proof-of-work/app/api/workouts/import/save/route.ts

@@ -15,6 +15,7 @@ const setSchema = z.object({
   calories: z.number().int().positive().optional(),
   watts: z.number().int().positive().optional(),
   rpe: z.number().int().min(1).max(10).optional(),
+  gear: z.number().int().min(1).max(5).optional(),
   notes: z.string().optional(),
 });
 
@@ -124,6 +125,7 @@ export async function POST(request: Request) {
             weight: set.weight || null,
             weightUnit: set.weightUnit || "lbs",
             rpe: set.rpe || null,
+            gear: set.gear || null,
             durationSeconds: set.durationSeconds || null,
             distance: set.distance || null,
             distanceUnit: set.distanceUnit || null,

proof-of-work/app/api/workouts/route.ts

@@ -26,6 +26,7 @@ const createWorkoutSchema = z.object({
         weight: z.number().positive().optional(),
         weightUnit: z.string().default("lbs"),
         rpe: z.number().int().min(1).max(10).optional(),
+        gear: z.number().int().min(1).max(5).optional(),
         durationSeconds: z.number().int().positive().optional(),
         distance: z.number().positive().optional(),
         distanceUnit: z.string().optional(),
@@ -175,6 +176,7 @@ export async function POST(request: NextRequest) {
                 weight: set.weight,
                 weightUnit: set.weightUnit,
                 rpe: set.rpe,
+                gear: set.gear,
                 durationSeconds: set.durationSeconds,
                 distance: set.distance,
                 distanceUnit: set.distanceUnit,

proof-of-work/app/main/ai/generate-workout/page.tsx

@@ -0,0 +1,70 @@
+import { redirect } from 'next/navigation';
+import Link from 'next/link';
+import { ChevronLeft } from 'lucide-react';
+import { getCurrentUser } from '@/lib/auth';
+import { prisma } from '@/lib/prisma';
+import GenerateWorkoutClient from '@/components/ai/GenerateWorkoutClient';
+
+export const dynamic = 'force-dynamic';
+
+export default async function GenerateWorkoutPage() {
+  const user = await getCurrentUser();
+  if (!user) redirect('/auth/login');
+
+  const [exercises, prefs, workoutCount] = await Promise.all([
+    prisma.exercise.findMany({
+      where: { userId: user.id },
+      select: { id: true, name: true, type: true },
+      orderBy: [{ type: 'asc' }, { name: 'asc' }],
+    }),
+    prisma.userPreferences.findUnique({
+      where: { userId: user.id },
+      select: { aiProvider: true, aiModel: true },
+    }),
+    prisma.workout.count({
+      where: { userId: user.id, deletedAt: null },
+    }),
+  ]);
+
+  const aiConfigured = !!prefs?.aiProvider && !!prefs?.aiModel;
+
+  return (
+    <div className="min-h-screen bg-[#0A0A0A]">
+      <div className="border-b border-zinc-800">
+        <div className="max-w-3xl mx-auto px-4 py-4 sm:py-6 flex items-center gap-3">
+          <Link
+            href="/main/ai"
+            className="text-zinc-400 hover:text-white"
+            aria-label="Back to AI"
+          >
+            <ChevronLeft className="w-5 h-5" />
+          </Link>
+          <h1 className="text-2xl sm:text-3xl font-bold text-white">
+            AI · Today&apos;s workout
+          </h1>
+        </div>
+      </div>
+      <div className="max-w-3xl mx-auto px-4 py-6">
+        {!aiConfigured ? (
+          <div className="bg-amber-950/30 border border-amber-900 rounded p-5 text-sm text-amber-200">
+            <p className="font-bold text-amber-100 mb-2">AI is not configured.</p>
+            <p>
+              Pick a provider, model, and (if needed) API key in{' '}
+              <Link href="/main/settings" className="underline hover:text-amber-100">
+                Settings → AI integration
+              </Link>{' '}
+              before you can generate a workout.
+            </p>
+          </div>
+        ) : (
+          <GenerateWorkoutClient
+            exercises={exercises}
+            providerLabel={prefs!.aiProvider!}
+            modelLabel={prefs!.aiModel!}
+            workoutCount={workoutCount}
+          />
+        )}
+      </div>
+    </div>
+  );
+}

proof-of-work/app/main/ai/history/[id]/page.tsx

@@ -34,7 +34,8 @@ export default async function GenerationDetailPage(props: {
 
   const [row, exercises] = await Promise.all([
     prisma.aIGeneration.findFirst({
-      where: { id: params.id, userId: user.id },
+      // Program history only — workout-kind rows aren't shown here.
+      where: { id: params.id, userId: user.id, kind: 'program' },
     }),
     prisma.exercise.findMany({
       where: { userId: user.id },

proof-of-work/app/main/ai/history/page.tsx

@@ -12,7 +12,10 @@ export default async function HistoryPage() {
   if (!user) redirect('/auth/login');
 
   const rows = await prisma.aIGeneration.findMany({
-    where: { userId: user.id },
+    // Program history only. Single-workout generations (kind="workout")
+    // are ephemeral — the durable record is the saved Workout — so they
+    // don't belong in this program-shaped list/detail.
+    where: { userId: user.id, kind: 'program' },
     orderBy: { createdAt: 'desc' },
     take: 25,
     select: {

proof-of-work/app/main/ai/page.tsx

@@ -1,6 +1,6 @@
 import { redirect } from 'next/navigation';
 import Link from 'next/link';
-import { Sparkles, ListChecks, History } from 'lucide-react';
+import { Sparkles, ListChecks, History, Dumbbell } from 'lucide-react';
 import { getCurrentUser } from '@/lib/auth';
 import { prisma } from '@/lib/prisma';
 
@@ -17,6 +17,14 @@ export default async function AIIndexPage() {
   const aiConfigured = !!prefs?.aiProvider && !!prefs?.aiModel;
 
   const cards = [
+    {
+      href: '/main/ai/generate-workout',
+      icon: Dumbbell,
+      title: "Today's workout",
+      blurb:
+        'Describe today\'s session in plain words and get a ready-to-log workout — exercises with suggested weights and reps from your history. Refine it, then pre-fill the log.',
+      disabled: !aiConfigured,
+    },
     {
       href: '/main/ai/generate',
       icon: Sparkles,

proof-of-work/app/main/import/page-csv.tsx

@@ -24,6 +24,7 @@ interface ParsedSet {
   calories?: number;
   watts?: number;
   rpe?: number;
+  gear?: number;
   customMetrics?: Record<string, string>;
   notes?: string;
 }
@@ -399,6 +400,9 @@ export default function ImportCSVPage() {
           if (typeof set.rpe === "number" && !Number.isNaN(set.rpe)) {
             payloadSet.rpe = set.rpe;
           }
+          if (typeof set.gear === "number" && !Number.isNaN(set.gear)) {
+            payloadSet.gear = set.gear;
+          }
           if (
             set.customMetrics &&
             typeof set.customMetrics === "object" &&
@@ -767,7 +771,7 @@ export default function ImportCSVPage() {
                 </p>
                 <p className="text-sm text-zinc-500">
                   CSV columns: date, exercise, set, weight, reps, duration_seconds,
-                  distance, distance_unit, calories, watts, rpe, notes, custom_*
+                  distance, distance_unit, calories, watts, rpe, gear, notes, custom_*
                 </p>
               </div>
               {loading && (
@@ -782,12 +786,12 @@ export default function ImportCSVPage() {
               CSV Format Example
             </h3>
             <pre className="text-xs text-zinc-400 overflow-x-auto bg-zinc-950 p-4 rounded border border-zinc-800">
-              {`date,exercise,set,weight,weight_unit,reps,duration_seconds,distance,distance_unit,calories,watts,rpe,notes,custom_temperature,custom_metrics_json
+              {`date,exercise,set,weight,weight_unit,reps,duration_seconds,distance,distance_unit,calories,watts,rpe,gear,notes,custom_temperature,custom_metrics_json
-2025-02-15,Bench,1,225,lbs,5,,,,,,8,good form,,
+2025-02-15,Bench,1,225,lbs,5,,,,,,8,,good form,,
-2025-02-15,Bench,2,225,lbs,5,,,,,,8,,,
+2025-02-15,Bench,2,225,lbs,5,,,,,,8,,,,
-2025-02-16,Squat,1,315,lbs,8,,,,,,9,30kg per leg,,
+2025-02-16,Squat,1,315,lbs,8,,,,,,9,,30kg per leg,,
-2025-02-17,Assault Bike,1,,, ,900,5,mi,120,157,7,,,"{\"resistance\":\"8\"}"
+2025-02-17,Assault Bike,1,,, ,900,5,mi,120,157,,4,,,"{\"resistance\":\"8\"}"
-2025-02-18,Cold Plunge,1,,, ,180,,,,,,felt great,50,`}
+2025-02-18,Cold Plunge,1,,, ,180,,,,,,,felt great,50,`}
             </pre>
           </div>
         </div>

proof-of-work/app/main/navigation.tsx

@@ -44,7 +44,8 @@ const navLinks: NavLink[] = [
     label: 'AI',
     icon: Sparkles,
     subItems: [
-      { href: '/main/ai/generate', label: 'Generate' },
+      { href: '/main/ai/generate-workout', label: "Today's workout" },
+      { href: '/main/ai/generate', label: 'Generate program' },
       { href: '/main/ai/history', label: 'History' },
       { href: '/main/ai/templates', label: 'Templates' },
     ],

proof-of-work/app/main/workouts/[id]/page.tsx

@@ -19,6 +19,7 @@ function buildSetSummary(set: {
   weightUnit?: string | null;
   reps?: number | null;
   rpe?: number | null;
+  gear?: number | null;
   notes?: string | null;
   durationSeconds?: number | null;
   distance?: number | null;
@@ -53,7 +54,8 @@ function buildSetSummary(set: {
       }
     } catch {}
   }
-  if (set.rpe) parts.push(`RPE ${set.rpe}`);
+  if (set.gear) parts.push(`Gear ${set.gear}`);
+  else if (set.rpe) parts.push(`RPE ${set.rpe}`);
   if (set.notes) parts.push(set.notes);
   return parts.length > 0 ? parts.join(" · ") : "No data";
 }

proof-of-work/app/main/workouts/new/page.tsx

@@ -5,6 +5,7 @@ import { getCurrentUser } from "@/lib/auth";
 import { getExercises } from "@/lib/db/exercises";
 import { getWorkoutById } from "@/lib/db/workouts";
 import WorkoutForm, { EditWorkoutData } from "@/components/workouts/WorkoutForm";
+import AiWorkoutPrefill from "@/components/workouts/AiWorkoutPrefill";
 
 export const metadata = {
   title: "Log Workout",
@@ -12,7 +13,7 @@ export const metadata = {
 };
 
 export default async function NewWorkoutPage(props: {
-  searchParams: Promise<{ edit?: string }>;
+  searchParams: Promise<{ edit?: string; from?: string }>;
 }) {
   const searchParams = await props.searchParams;
   const user = await getCurrentUser();
@@ -22,6 +23,11 @@ export default async function NewWorkoutPage(props: {
 
   const exercises = await getExercises(user.id);
 
+  // Coming from the AI "today's workout" flow: the suggestion is in
+  // sessionStorage (client-only), so a client wrapper reads it and
+  // pre-fills the form. No ?edit fetch here.
+  const fromAi = searchParams.from === "ai";
+
   // If ?edit=WORKOUT_ID, fetch existing workout for editing
   let editWorkout: EditWorkoutData | undefined;
   if (searchParams.edit) {
@@ -50,6 +56,7 @@ export default async function NewWorkoutPage(props: {
           reps: set.reps ?? undefined,
           weight: set.weight ?? undefined,
           rpe: set.rpe ?? undefined,
+          gear: set.gear ?? undefined,
           durationSeconds: set.durationSeconds ?? undefined,
           distance: set.distance ?? undefined,
           calories: set.calories ?? undefined,
@@ -94,11 +101,15 @@ export default async function NewWorkoutPage(props: {
 
       {/* Form */}
       <div className="max-w-2xl mx-auto px-4 py-6 pb-12">
-        <WorkoutForm
+        {fromAi ? (
-          exercises={exercises}
+          <AiWorkoutPrefill exercises={exercises} />
-          recentlyUsedExercises={[]}
+        ) : (
-          editWorkout={editWorkout}
+          <WorkoutForm
-        />
+            exercises={exercises}
+            recentlyUsedExercises={[]}
+            editWorkout={editWorkout}
+          />
+        )}
       </div>
     </div>
   );

proof-of-work/components/ai/GenerateWorkoutClient.tsx

@@ -0,0 +1,626 @@
+'use client';
+
+import { useEffect, useMemo, useRef, useState } from 'react';
+import { useRouter } from 'next/navigation';
+import { Loader2, Sparkles } from 'lucide-react';
+import { lenientJsonParse } from '@/lib/ai/lenientJson';
+import { estimateCost, formatCost } from '@/lib/ai/pricing';
+import type { AiWorkoutDraft } from '@/lib/ai/workoutDraft';
+
+interface LibraryExercise {
+  id: string;
+  name: string;
+  type: string;
+}
+
+// AI output shape — mirrors lib/ai/workoutSchema.ts (AIWorkout).
+interface AIWorkoutExercise {
+  exerciseId: string | null;
+  exerciseName: string;
+  order: number;
+  sets?: number | null;
+  reps?: number | null;
+  suggestedWeight?: number | null;
+  suggestedWeightUnit?: 'lbs' | 'kg' | null;
+  rpe?: number | null;
+  gear?: number | null;
+  durationSeconds?: number | null;
+  notes?: string | null;
+}
+interface AIWorkout {
+  name: string;
+  notes?: string | null;
+  exercises: AIWorkoutExercise[];
+}
+
+// The ephemeral draft we hand to the New Workout form via sessionStorage.
+export const AI_WORKOUT_DRAFT_KEY = 'ai-workout-draft';
+
+type Phase =
+  | { kind: 'idle' }
+  | { kind: 'streaming'; raw: string; lastPartial: Partial<AIWorkout> | null }
+  | { kind: 'failed'; raw: string; message: string };
+
+export default function GenerateWorkoutClient({
+  exercises,
+  providerLabel,
+  modelLabel,
+  workoutCount,
+}: {
+  exercises: LibraryExercise[];
+  providerLabel: string;
+  modelLabel: string;
+  workoutCount: number;
+}) {
+  const router = useRouter();
+  const [userInput, setUserInput] = useState('');
+  const [includeHistory, setIncludeHistory] = useState(workoutCount >= 1);
+  const [phase, setPhase] = useState<Phase>({ kind: 'idle' });
+  // The editable suggestion once parsed. Lifted to the parent so the
+  // Refine action can send the user's current edits back as the prior
+  // workout. null until the first successful parse.
+  const [workout, setWorkout] = useState<AIWorkout | null>(null);
+  // Refine instruction lives here (not in WorkoutPreview) because the
+  // preview unmounts while streaming — keeping it in the parent means a
+  // failed refine doesn't lose what the user typed; we clear it only on
+  // a successful regeneration.
+  const [refineInput, setRefineInput] = useState('');
+  const [tokens, setTokens] = useState<{ in?: number; out?: number; durationMs?: number }>({});
+  const closeStreamRef = useRef<(() => void) | null>(null);
+
+  const streaming = phase.kind === 'streaming';
+
+  /**
+   * Run a generation. `priorWorkout` present → REVISION mode: `input`
+   * is the change instruction and the model re-emits the full workout.
+   */
+  const runGeneration = async (input: string, priorWorkout?: AIWorkout) => {
+    if (!input.trim()) return;
+    setPhase({ kind: 'streaming', raw: '', lastPartial: null });
+    setTokens({});
+
+    let id: string;
+    try {
+      const res = await fetch('/api/ai/generate-workout', {
+        method: 'POST',
+        headers: { 'content-type': 'application/json' },
+        body: JSON.stringify({
+          userInput: input,
+          includeHistory,
+          priorWorkout: priorWorkout ?? null,
+        }),
+      });
+      const body = await res.json().catch(() => ({}));
+      if (!res.ok) {
+        setPhase({ kind: 'failed', raw: '', message: body.error ?? `HTTP ${res.status}` });
+        return;
+      }
+      id = body.id;
+    } catch (e) {
+      setPhase({ kind: 'failed', raw: '', message: (e as Error).message });
+      return;
+    }
+
+    attachStream(id);
+  };
+
+  const attachStream = (id: string) => {
+    const es = new EventSource(`/api/ai/generations/${id}/stream`);
+    closeStreamRef.current = () => es.close();
+    let raw = '';
+    let lastPartial: Partial<AIWorkout> | null = null;
+
+    es.addEventListener('text', (ev) => {
+      const data = JSON.parse((ev as MessageEvent).data);
+      raw += data.delta;
+      const next = lenientJsonParse(raw) as Partial<AIWorkout> | null;
+      if (next) lastPartial = next; // sticky — kills flicker between parses
+      setPhase({ kind: 'streaming', raw, lastPartial });
+    });
+    es.addEventListener('usage', (ev) => {
+      const data = JSON.parse((ev as MessageEvent).data);
+      setTokens((t) => ({ ...t, in: data.tokensIn, out: data.tokensOut }));
+    });
+    es.addEventListener('complete', async (ev) => {
+      const data = JSON.parse((ev as MessageEvent).data);
+      es.close();
+      closeStreamRef.current = null;
+      setTokens((t) => ({
+        ...t,
+        in: data.tokensIn ?? t.in,
+        out: data.tokensOut ?? t.out,
+        durationMs: data.durationMs,
+      }));
+      if (data.parsedOk) {
+        const r = await fetch(`/api/ai/generations/${id}`);
+        if (r.ok) {
+          const gen = await r.json();
+          if (gen.parsedProgram) {
+            setWorkout(JSON.parse(gen.parsedProgram) as AIWorkout);
+            setRefineInput(''); // consumed — clear only on success
+            setPhase({ kind: 'idle' });
+            return;
+          }
+        }
+      }
+      setPhase({
+        kind: 'failed',
+        raw,
+        message: data.errorMessage ?? 'Failed to parse model output.',
+      });
+    });
+    es.onerror = () => {
+      if (es.readyState === EventSource.CLOSED) {
+        closeStreamRef.current = null;
+        setPhase((p) =>
+          p.kind === 'streaming'
+            ? {
+                kind: 'failed',
+                raw: p.raw,
+                message:
+                  'Stream disconnected. The generation may still be running — check AI · History.',
+              }
+            : p,
+        );
+      }
+    };
+  };
+
+  // Warn before unload while streaming (the runner keeps going server-side).
+  useEffect(() => {
+    if (!streaming) return;
+    const onBeforeUnload = (e: BeforeUnloadEvent) => {
+      e.preventDefault();
+      e.returnValue = '';
+    };
+    window.addEventListener('beforeunload', onBeforeUnload);
+    return () => window.removeEventListener('beforeunload', onBeforeUnload);
+  }, [streaming]);
+
+  // Detach on unmount; the server keeps generating regardless.
+  useEffect(() => () => closeStreamRef.current?.(), []);
+
+  const costStr = useMemo(() => {
+    if (tokens.in == null || tokens.out == null) return null;
+    return formatCost(
+      estimateCost({
+        provider: providerLabel,
+        model: modelLabel,
+        tokensIn: tokens.in,
+        tokensOut: tokens.out,
+      }),
+    );
+  }, [providerLabel, modelLabel, tokens.in, tokens.out]);
+
+  const showResult = streaming || phase.kind === 'failed' || workout != null;
+
+  return (
+    <div className="space-y-6">
+      <div className="text-xs text-zinc-500 uppercase tracking-wider">
+        Provider: <span className="text-zinc-300">{providerLabel}</span>
+        {' · '}Model: <span className="text-zinc-300">{modelLabel}</span>
+      </div>
+
+      <section className="bg-zinc-900 border border-zinc-800 rounded p-4 space-y-4">
+        <Field label="Describe today's workout">
+          <textarea
+            value={userInput}
+            onChange={(e) => setUserInput(e.target.value)}
+            placeholder="e.g. Upper body, focus on shoulders. Overhead press, 4 working sets. Abs: landmine rotation, oblique cable, landmine hollow-body hold. Pull-ups, biceps and triceps."
+            rows={6}
+            className={inputClass}
+            disabled={streaming}
+          />
+        </Field>
+
+        <label className="flex items-start gap-2 text-xs text-zinc-300">
+          <input
+            type="checkbox"
+            checked={includeHistory}
+            onChange={(e) => setIncludeHistory(e.target.checked)}
+            disabled={streaming || workoutCount === 0}
+            className="mt-0.5"
+          />
+          <span>
+            Use my history to suggest weights{' '}
+            <span className="text-zinc-500">
+              ({workoutCount === 0
+                ? 'no workouts logged yet — disabled'
+                : 'last 90 days · recent working weights per exercise'}
+              )
+            </span>
+          </span>
+        </label>
+
+        <button
+          type="button"
+          onClick={() => runGeneration(userInput)}
+          disabled={!userInput.trim() || streaming}
+          className="inline-flex items-center gap-2 px-5 py-2 rounded bg-white text-black font-bold text-xs uppercase tracking-wider hover:bg-gray-100 disabled:bg-zinc-700 disabled:text-zinc-500"
+        >
+          <Sparkles className="w-4 h-4" />
+          {workout ? 'Regenerate' : 'Generate workout'}
+        </button>
+      </section>
+
+      {showResult && (
+        <section className="space-y-3">
+          <div className="flex items-center justify-between">
+            <h2 className="text-sm font-semibold text-white uppercase tracking-wider">
+              {streaming ? 'Generating…' : 'Suggested workout'}
+            </h2>
+            <span className="text-[11px] text-zinc-500 uppercase tracking-wider">
+              {tokens.in != null && (
+                <>
+                  {tokens.in} in · {tokens.out ?? '?'} out
+                </>
+              )}
+              {costStr && <> · {costStr}</>}
+              {tokens.durationMs != null && (
+                <> · {(tokens.durationMs / 1000).toFixed(1)}s</>
+              )}
+            </span>
+          </div>
+
+          {streaming && (
+            <>
+              {phase.lastPartial ? (
+                <PartialPreview partial={phase.lastPartial} />
+              ) : (
+                <div className="text-xs text-zinc-500 italic flex items-center gap-2">
+                  <Loader2 className="w-3 h-3 animate-spin" />
+                  Waiting for the first parseable JSON…
+                </div>
+              )}
+              <details className="text-xs text-zinc-500">
+                <summary className="cursor-pointer">Raw stream</summary>
+                <div className="bg-zinc-950 border border-zinc-800 rounded p-3 font-mono text-[11px] text-zinc-400 max-h-80 overflow-auto whitespace-pre-wrap mt-2">
+                  {phase.raw || '(waiting for first token…)'}
+                  <Loader2 className="inline w-3 h-3 animate-spin ml-2" />
+                </div>
+              </details>
+            </>
+          )}
+
+          {phase.kind === 'failed' && (
+            <>
+              <div className="bg-red-950/40 border border-red-900 rounded p-3 text-sm text-red-300">
+                {phase.message}
+              </div>
+              {phase.raw && (
+                <details className="text-xs text-zinc-500">
+                  <summary className="cursor-pointer">Raw response</summary>
+                  <pre className="bg-zinc-950 border border-zinc-800 rounded p-3 mt-2 whitespace-pre-wrap">
+                    {phase.raw}
+                  </pre>
+                </details>
+              )}
+            </>
+          )}
+
+          {!streaming && workout && (
+            <WorkoutPreview
+              workout={workout}
+              setWorkout={(updater) => setWorkout((w) => (w ? updater(w) : w))}
+              exercises={exercises}
+              refineInput={refineInput}
+              setRefineInput={setRefineInput}
+              onRefine={() => runGeneration(refineInput, workout)}
+              onUse={(draft) => {
+                sessionStorage.setItem(AI_WORKOUT_DRAFT_KEY, JSON.stringify(draft));
+                router.push('/main/workouts/new?from=ai');
+              }}
+            />
+          )}
+        </section>
+      )}
+    </div>
+  );
+}
+
+function WorkoutPreview({
+  workout,
+  setWorkout,
+  exercises,
+  refineInput,
+  setRefineInput,
+  onRefine,
+  onUse,
+}: {
+  workout: AIWorkout;
+  setWorkout: (updater: (w: AIWorkout) => AIWorkout) => void;
+  exercises: LibraryExercise[];
+  refineInput: string;
+  setRefineInput: (v: string) => void;
+  onRefine: () => void;
+  onUse: (draft: AiWorkoutDraft) => void;
+}) {
+  const [error, setError] = useState<string | null>(null);
+
+  const exerciseLookup = useMemo(
+    () => new Map(exercises.map((e) => [e.id, e])),
+    [exercises],
+  );
+
+  const unresolvedCount = useMemo(
+    () =>
+      workout.exercises.filter(
+        (ex) => !ex.exerciseId || !exerciseLookup.has(ex.exerciseId),
+      ).length,
+    [workout, exerciseLookup],
+  );
+
+  const updateExercise = (
+    idx: number,
+    patch: Partial<AIWorkoutExercise>,
+  ) => {
+    setWorkout((w) => {
+      const next = structuredClone(w);
+      next.exercises[idx] = { ...next.exercises[idx], ...patch };
+      return next;
+    });
+  };
+
+  const removeExercise = (idx: number) => {
+    setWorkout((w) => {
+      const next = structuredClone(w);
+      next.exercises.splice(idx, 1);
+      next.exercises.forEach((ex, i) => (ex.order = i));
+      return next;
+    });
+  };
+
+  const numOrNull = (v: string) => {
+    if (v.trim() === '') return null;
+    const n = Number(v);
+    return Number.isFinite(n) ? n : null;
+  };
+
+  const handleUse = () => {
+    if (unresolvedCount > 0) {
+      setError(`Map or remove the ${unresolvedCount} unknown exercise(s) first.`);
+      return;
+    }
+    setError(null);
+    onUse({
+      name: workout.name,
+      notes: workout.notes ?? undefined,
+      exercises: workout.exercises.map((ex) => ({
+        exerciseId: ex.exerciseId!,
+        sets: ex.sets && ex.sets > 0 ? ex.sets : 3,
+        reps: ex.reps ?? undefined,
+        suggestedWeight: ex.suggestedWeight ?? undefined,
+        suggestedWeightUnit: ex.suggestedWeightUnit ?? undefined,
+        rpe: ex.rpe ?? undefined,
+        gear: ex.gear ?? undefined,
+        durationSeconds: ex.durationSeconds ?? undefined,
+        notes: ex.notes ?? undefined,
+      })),
+    });
+  };
+
+  return (
+    <div className="bg-zinc-900 border border-zinc-800 rounded p-4 space-y-4">
+      <div>
+        <input
+          value={workout.name}
+          onChange={(e) => setWorkout((w) => ({ ...w, name: e.target.value }))}
+          className="text-lg font-bold text-white bg-transparent border-b border-transparent hover:border-zinc-700 focus:border-zinc-500 focus:outline-none w-full"
+        />
+        {workout.notes && (
+          <p className="text-sm text-zinc-400 mt-1 italic">{workout.notes}</p>
+        )}
+        <p className="text-xs text-zinc-500 mt-1">
+          {workout.exercises.length} exercise
+          {workout.exercises.length === 1 ? '' : 's'}
+        </p>
+      </div>
+
+      {unresolvedCount > 0 && (
+        <div className="rounded bg-amber-950/30 border border-amber-900 px-3 py-2 text-xs text-amber-200">
+          {unresolvedCount} exercise(s) the AI couldn&apos;t map to your library.
+          Pick a replacement or remove them before using this workout.
+        </div>
+      )}
+
+      <ul className="space-y-2">
+        {workout.exercises.map((ex, idx) => {
+          const isUnknown = !ex.exerciseId || !exerciseLookup.has(ex.exerciseId);
+          const lib = ex.exerciseId ? exerciseLookup.get(ex.exerciseId) : null;
+          return (
+            <li
+              key={idx}
+              className={`rounded p-3 ${
+                isUnknown
+                  ? 'bg-amber-950/30 border border-amber-900'
+                  : 'bg-zinc-950 border border-zinc-800'
+              }`}
+            >
+              <div className="flex items-start justify-between gap-2">
+                <div className="text-white text-sm">
+                  {lib?.name ?? ex.exerciseName}
+                  {isUnknown && (
+                    <span className="ml-2 text-[10px] uppercase tracking-wider text-amber-400">
+                      not in library
+                    </span>
+                  )}
+                </div>
+                <button
+                  type="button"
+                  onClick={() => removeExercise(idx)}
+                  className="text-xs text-red-400 hover:text-red-300 px-1"
+                  title="Remove from workout"
+                >
+                  ✕
+                </button>
+              </div>
+
+              {isUnknown && (
+                <select
+                  value={ex.exerciseId ?? ''}
+                  onChange={(e) =>
+                    updateExercise(idx, { exerciseId: e.target.value || null })
+                  }
+                  className="mt-2 w-full text-xs px-2 py-1 rounded border border-amber-900 bg-zinc-900 text-white"
+                >
+                  <option value="">Map to existing exercise…</option>
+                  {exercises.map((opt) => (
+                    <option key={opt.id} value={opt.id}>
+                      {opt.name} ({opt.type})
+                    </option>
+                  ))}
+                </select>
+              )}
+
+              <div className="mt-2 grid grid-cols-3 gap-2">
+                <NumField
+                  label="Sets"
+                  value={ex.sets}
+                  onChange={(v) => updateExercise(idx, { sets: v })}
+                />
+                <NumField
+                  label="Reps"
+                  value={ex.reps}
+                  onChange={(v) => updateExercise(idx, { reps: v })}
+                />
+                <NumField
+                  label={`Weight${ex.suggestedWeightUnit ? ` (${ex.suggestedWeightUnit})` : ''}`}
+                  value={ex.suggestedWeight}
+                  step="any"
+                  onChange={(v) => updateExercise(idx, { suggestedWeight: v })}
+                />
+              </div>
+              {ex.notes && (
+                <div className="text-xs text-zinc-400 mt-2 italic">{ex.notes}</div>
+              )}
+            </li>
+          );
+        })}
+      </ul>
+
+      <div className="border-t border-zinc-800 pt-4 space-y-3">
+        <Field label="Refine (send a change back to the AI)">
+          <div className="flex gap-2">
+            <input
+              value={refineInput}
+              onChange={(e) => setRefineInput(e.target.value)}
+              onKeyDown={(e) => {
+                // Cleared by the parent only on a successful regeneration, so
+                // a failed refine keeps what the user typed.
+                if (e.key === 'Enter' && refineInput.trim()) onRefine();
+              }}
+              placeholder="e.g. make overhead press 5 sets; swap the oblique exercise"
+              className={inputClass}
+            />
+            <button
+              type="button"
+              onClick={() => {
+                if (refineInput.trim()) onRefine();
+              }}
+              disabled={!refineInput.trim()}
+              className="shrink-0 px-4 py-2 rounded bg-zinc-700 text-white font-bold text-xs uppercase tracking-wider hover:bg-zinc-600 disabled:bg-zinc-800 disabled:text-zinc-600"
+            >
+              Refine
+            </button>
+          </div>
+        </Field>
+
+        {error && (
+          <div className="rounded bg-red-900/50 px-3 py-2 border border-red-800 text-xs text-red-400">
+            {error}
+          </div>
+        )}
+
+        <button
+          type="button"
+          onClick={handleUse}
+          disabled={unresolvedCount > 0 || workout.exercises.length === 0}
+          className="px-5 py-2 rounded bg-emerald-700 text-white font-bold text-xs uppercase tracking-wider hover:bg-emerald-600 disabled:bg-zinc-700 disabled:text-zinc-500"
+        >
+          Use this workout
+        </button>
+        <p className="text-[11px] text-zinc-500">
+          Opens a pre-filled workout — nothing is saved until you save it there.
+        </p>
+      </div>
+    </div>
+  );
+
+  function NumField({
+    label,
+    value,
+    step,
+    onChange,
+  }: {
+    label: string;
+    value?: number | null;
+    step?: string;
+    onChange: (v: number | null) => void;
+  }) {
+    return (
+      <label className="block">
+        <span className="text-[10px] font-semibold text-zinc-500 uppercase tracking-wider block mb-0.5">
+          {label}
+        </span>
+        <input
+          type="number"
+          inputMode="decimal"
+          step={step}
+          value={value ?? ''}
+          onChange={(e) => onChange(numOrNull(e.target.value))}
+          className="w-full px-2 py-1 text-sm rounded border border-zinc-700 bg-zinc-800 text-white focus:outline-none focus:ring-1 focus:ring-white/30"
+        />
+      </label>
+    );
+  }
+}
+
+const inputClass =
+  'w-full px-3 py-2 text-sm rounded border border-zinc-700 bg-zinc-800 text-white placeholder:text-zinc-500 focus:outline-none focus:ring-2 focus:ring-white/30';
+
+function Field({ label, children }: { label: string; children: React.ReactNode }) {
+  return (
+    <label className="block">
+      <span className="text-[11px] font-semibold text-zinc-400 uppercase tracking-wider block mb-1">
+        {label}
+      </span>
+      {children}
+    </label>
+  );
+}
+
+function PartialPreview({ partial }: { partial: Partial<AIWorkout> }) {
+  const exercises = (partial.exercises as AIWorkoutExercise[] | undefined) ?? [];
+  return (
+    <div className="bg-zinc-950 border border-zinc-800 rounded p-3 space-y-2">
+      <div className="flex items-center gap-2 text-xs">
+        <Loader2 className="w-3 h-3 animate-spin text-zinc-500" />
+        <span className="text-zinc-400">
+          Building workout…{' '}
+          {partial.name && (
+            <span className="text-white font-semibold">{partial.name}</span>
+          )}
+        </span>
+      </div>
+      {exercises.length > 0 && (
+        <ul className="text-xs text-zinc-300 space-y-1">
+          {exercises.map((ex, i) => (
+            <li key={i}>
+              <span className="text-zinc-500">{(ex?.order ?? i) + 1}.</span>{' '}
+              {ex?.exerciseName ?? '…'}
+              {ex?.sets ? (
+                <span className="text-zinc-500">
+                  {' '}
+                  · {ex.sets}×{ex.reps ?? '?'}
+                  {ex.suggestedWeight != null
+                    ? ` @ ${ex.suggestedWeight}${ex.suggestedWeightUnit ?? ''}`
+                    : ''}
+                </span>
+              ) : null}
+            </li>
+          ))}
+        </ul>
+      )}
+    </div>
+  );
+}

proof-of-work/components/workouts/AiWorkoutPrefill.tsx

@@ -0,0 +1,70 @@
+'use client';
+
+import { useEffect, useState } from 'react';
+import { Exercise } from '@prisma/client';
+import WorkoutForm, { EditWorkoutData } from './WorkoutForm';
+import { AI_WORKOUT_DRAFT_KEY } from '@/components/ai/GenerateWorkoutClient';
+import { buildPrefillExercises, type AiWorkoutDraft } from '@/lib/ai/workoutDraft';
+
+/**
+ * Reads the ephemeral AI workout draft from sessionStorage (stashed by
+ * GenerateWorkoutClient before navigating to /main/workouts/new?from=ai),
+ * expands each suggested exercise into N pre-filled SetLogs, and renders
+ * the normal WorkoutForm. Nothing is persisted until the user saves
+ * through the regular workout path.
+ *
+ * The draft has no workout id, so WorkoutForm's first save CREATEs.
+ * Effort follows the app convention: cardio → gear (1-5), else → rpe.
+ *
+ * If the draft is missing (e.g. a refresh cleared it), we fall back to a
+ * blank form so the page is never broken.
+ */
+export default function AiWorkoutPrefill({
+  exercises,
+}: {
+  exercises: Exercise[];
+}) {
+  // Read + build once on mount via a lazy initializer. This stays PURE
+  // (no sessionStorage mutation) so React's StrictMode double-invoke is
+  // safe — both passes read the same draft. The one-shot removal happens
+  // in the effect below, after the value is captured.
+  const [editWorkout] = useState<EditWorkoutData | undefined>(() => {
+    if (typeof window === 'undefined') return undefined;
+    const raw = sessionStorage.getItem(AI_WORKOUT_DRAFT_KEY);
+    if (!raw) return undefined;
+
+    let draft: AiWorkoutDraft;
+    try {
+      draft = JSON.parse(raw);
+    } catch {
+      return undefined;
+    }
+
+    const builtExercises: EditWorkoutData['exercises'] =
+      buildPrefillExercises(draft, exercises);
+
+    if (builtExercises.length === 0) return undefined;
+
+    return {
+      // No id → first save CREATEs a new workout.
+      name: draft.name || '',
+      date: new Date().toISOString(),
+      notes: draft.notes,
+      exercises: builtExercises,
+    };
+  });
+
+  // Clear the one-shot draft after mount so a manual reload starts blank.
+  // removeItem is idempotent, so StrictMode's double-run is harmless.
+  useEffect(() => {
+    sessionStorage.removeItem(AI_WORKOUT_DRAFT_KEY);
+  }, []);
+
+  return (
+    <WorkoutForm
+      exercises={exercises}
+      recentlyUsedExercises={[]}
+      editWorkout={editWorkout}
+    />
+  );
+}

proof-of-work/components/workouts/SetRow.tsx

@@ -18,9 +18,12 @@ export interface SetRowProps {
   setNumber: number;
   inputFields?: InputField[];
   weightUnit?: string;
+  /** Cardio sets log breathing "Gear" (1-5) instead of RPE (6-10). */
+  isCardio?: boolean;
   initialReps?: number;
   initialWeight?: number;
   initialRpe?: number;
+  initialGear?: number;
   initialNotes?: string;
   initialDuration?: number;
   initialDistance?: number;
@@ -33,6 +36,7 @@ export interface SetRowProps {
     reps?: number;
     weight?: number;
     rpe?: number;
+    gear?: number;
     notes?: string;
     durationSeconds?: number;
     distance?: number;
@@ -45,6 +49,7 @@ export interface SetRowProps {
     weight?: string;
     reps?: string;
     rpe?: string;
+    gear?: string;
     notes?: string;
     duration?: string;
     distance?: string;
@@ -58,9 +63,11 @@ export default function SetRow({
   setNumber,
   inputFields = ["sets", "reps", "weight"],
   weightUnit = "lbs",
+  isCardio = false,
   initialReps,
   initialWeight,
   initialRpe,
+  initialGear,
   initialNotes,
   initialDuration,
   initialDistance,
@@ -91,6 +98,7 @@ export default function SetRow({
   const [reps, setReps] = useState(initialReps?.toString() || "");
   const [weight, setWeight] = useState(initialWeight?.toString() || "");
   const [rpe, setRpe] = useState(initialRpe?.toString() || "");
+  const [gear, setGear] = useState(initialGear?.toString() || "");
   const [notes, setNotes] = useState(initialNotes || "");
   const [duration, setDuration] = useState(secondsToMinuteString(initialDuration));
   const [distance, setDistance] = useState(initialDistance?.toString() || "");
@@ -134,6 +142,7 @@ export default function SetRow({
       reps?: string;
       weight?: string;
       rpe?: string;
+      gear?: string;
       notes?: string;
       duration?: string;
       distance?: string;
@@ -144,6 +153,7 @@ export default function SetRow({
       const r = overrides.reps ?? reps;
       const w = overrides.weight ?? weight;
       const p = overrides.rpe ?? rpe;
+      const gr = overrides.gear ?? gear;
       const n = overrides.notes ?? notes;
       const dur = overrides.duration ?? duration;
       const dist = overrides.distance ?? distance;
@@ -158,6 +168,7 @@ export default function SetRow({
         reps: r ? parseInt(r) : undefined,
         weight: w ? parseFloat(w) : undefined,
         rpe: p ? parseInt(p) : undefined,
+        gear: gr ? parseInt(gr) : undefined,
         notes: n || undefined,
         durationSeconds: minuteStringToSeconds(dur),
         distance: dist ? parseFloat(dist) : undefined,
@@ -169,7 +180,7 @@ export default function SetRow({
             : undefined,
       });
     },
-    [reps, weight, rpe, notes, duration, distance, calories, watts, customValues, onUpdate]
+    [reps, weight, rpe, gear, notes, duration, distance, calories, watts, customValues, onUpdate]
   );
 
   const handleConfirm = () => {
@@ -192,7 +203,7 @@ export default function SetRow({
   const handleNextSet = () => {
     emitUpdate({});
     setLocked(true);
-    onNextSet?.({ weight, reps, rpe, notes, duration, distance, calories, watts });
+    onNextSet?.({ weight, reps, rpe, gear, notes, duration, distance, calories, watts });
   };
 
   // Build a summary string for the locked view
@@ -208,7 +219,11 @@ export default function SetRow({
       const value = customValues[field];
       if (value) parts.push(`${field}: ${value}`);
     }
-    if (rpe) parts.push(`RPE ${rpe}`);
+    if (isCardio) {
+      if (gear) parts.push(`Gear ${gear}`);
+    } else if (rpe) {
+      parts.push(`RPE ${rpe}`);
+    }
     if (showNotesField && notes) parts.push(notes);
     return parts.length > 0 ? parts.join(" · ") : "No data";
   };
@@ -396,28 +411,52 @@ export default function SetRow({
           </div>
         )}
 
-        {/* RPE select — always shown */}
+        {/* Effort select — Gear (1-5, breathing gear) for cardio, else RPE (6-10) */}
-        <div className="flex-1 min-w-[50px] max-w-[60px]">
+        {isCardio ? (
-          <label className="block text-[10px] font-medium text-zinc-500 mb-0.5">
+          <div className="flex-1 min-w-[50px] max-w-[60px]">
-            RPE
+            <label className="block text-[10px] font-medium text-zinc-500 mb-0.5">
-          </label>
+              Gear
-          <select
+            </label>
-            value={rpe}
+            <select
-            onChange={(e) => {
+              value={gear}
-              const val = e.target.value;
+              onChange={(e) => {
-              setRpe(val);
+                const val = e.target.value;
-              emitUpdate({ rpe: val });
+                setGear(val);
-            }}
+                emitUpdate({ gear: val });
-            className="w-full px-1.5 py-1.5 border border-zinc-700 rounded-md text-sm bg-zinc-800 text-white focus:outline-none focus:ring-2 focus:ring-white/20"
+              }}
-          >
+              className="w-full px-1.5 py-1.5 border border-zinc-700 rounded-md text-sm bg-zinc-800 text-white focus:outline-none focus:ring-2 focus:ring-white/20"
-            <option value="">-</option>
+            >
-            <option value="6">6</option>
+              <option value="">-</option>
-            <option value="7">7</option>
+              <option value="1">1</option>
-            <option value="8">8</option>
+              <option value="2">2</option>
-            <option value="9">9</option>
+              <option value="3">3</option>
-            <option value="10">10</option>
+              <option value="4">4</option>
-          </select>
+              <option value="5">5</option>
-        </div>
+            </select>
+          </div>
+        ) : (
+          <div className="flex-1 min-w-[50px] max-w-[60px]">
+            <label className="block text-[10px] font-medium text-zinc-500 mb-0.5">
+              RPE
+            </label>
+            <select
+              value={rpe}
+              onChange={(e) => {
+                const val = e.target.value;
+                setRpe(val);
+                emitUpdate({ rpe: val });
+              }}
+              className="w-full px-1.5 py-1.5 border border-zinc-700 rounded-md text-sm bg-zinc-800 text-white focus:outline-none focus:ring-2 focus:ring-white/20"
+            >
+              <option value="">-</option>
+              <option value="6">6</option>
+              <option value="7">7</option>
+              <option value="8">8</option>
+              <option value="9">9</option>
+              <option value="10">10</option>
+            </select>
+          </div>
+        )}
 
         {/* Next set button — confirm + add new pre-filled set */}
         {onNextSet && (

proof-of-work/components/workouts/WorkoutForm.tsx

@@ -7,6 +7,7 @@ import { ChevronDown, ChevronUp, Loader, Trash2, Plus, Save, Pencil, Check, Arro
 import ExercisePicker from "./ExercisePicker";
 import SetRow, { InputField } from "./SetRow";
 import { formatSetsSummary } from "@/lib/formatSets";
+import { isCardioExercise } from "@/lib/exerciseOptions";
 
 // --------------- Exercise History Popup ---------------
 type HistoryEntry = {
@@ -232,6 +233,7 @@ interface ExerciseWithSets {
     reps?: number;
     weight?: number;
     rpe?: number;
+    gear?: number;
     durationSeconds?: number;
     distance?: number;
     calories?: number;
@@ -243,7 +245,10 @@ interface ExerciseWithSets {
 }
 
 export interface EditWorkoutData {
-  id: string;
+  /// Existing workout id when editing. Omitted for a pre-filled NEW
+  /// workout (e.g. an AI suggestion) so `savedWorkoutId` starts null and
+  /// the first save CREATEs instead of PATCHing a nonexistent id.
+  id?: string;
   name: string;
   date: string; // ISO string
   durationMinutes?: number | null;
@@ -257,6 +262,7 @@ export interface EditWorkoutData {
       reps?: number;
       weight?: number;
       rpe?: number;
+      gear?: number;
       durationSeconds?: number;
       distance?: number;
       calories?: number;
@@ -344,6 +350,7 @@ export default function WorkoutForm({
             weight: s.weight,
             weightUnit: (e.exercise as any).defaultWeightUnit || "lbs",
             rpe: s.rpe,
+            gear: s.gear,
             durationSeconds: s.durationSeconds,
             distance: s.distance,
             distanceUnit: s.distance !== undefined ? "mi" : undefined,
@@ -507,6 +514,7 @@ export default function WorkoutForm({
       reps?: number;
       weight?: number;
       rpe?: number;
+      gear?: number;
       notes?: string;
       durationSeconds?: number;
       distance?: number;
@@ -559,6 +567,7 @@ export default function WorkoutForm({
       weight?: string;
       reps?: string;
       rpe?: string;
+      gear?: string;
       notes?: string;
       duration?: string;
       distance?: string;
@@ -580,6 +589,7 @@ export default function WorkoutForm({
                 weight: currentValues.weight ? parseFloat(currentValues.weight) : undefined,
                 reps: undefined, // User typically changes reps per set
                 rpe: currentValues.rpe ? parseInt(currentValues.rpe) : undefined,
+                gear: currentValues.gear ? parseInt(currentValues.gear) : undefined,
                 notes: currentValues.notes || undefined,
                 forceEdit: true, // Start in edit mode even though weight is pre-filled
               },
@@ -648,9 +658,11 @@ export default function WorkoutForm({
         if (!response.ok) throw new Error("Failed to save workout");
       }
 
-      // Navigate back: to detail page if editing, otherwise to list
+      // Navigate back: to detail page if we have a workout id (editing, or
-      if (editWorkout) {
+      // an AI-prefilled workout that has now been created), otherwise to list.
-        router.push(`/main/workouts/${savedWorkoutId || editWorkout.id}`);
+      const detailId = savedWorkoutId || editWorkout?.id;
+      if (detailId) {
+        router.push(`/main/workouts/${detailId}`);
       } else {
         router.push("/main/workouts");
       }
@@ -856,9 +868,11 @@ export default function WorkoutForm({
                           setNumber={set.setNumber}
                           inputFields={parseInputFields(item.exercise)}
                           weightUnit={(item.exercise as any).defaultWeightUnit || "lbs"}
+                          isCardio={isCardioExercise(item.exercise)}
                           initialReps={set.reps}
                           initialWeight={set.weight}
                           initialRpe={set.rpe}
+                          initialGear={set.gear}
                           initialDuration={set.durationSeconds}
                           initialDistance={set.distance}
                           initialCalories={set.calories}

proof-of-work/lib/ai/generationRunner.ts

@@ -34,6 +34,7 @@
 import type { PrismaClient } from '@prisma/client';
 import { getProvider } from './providers';
 import { parseAIProgram } from './programSchema';
+import { parseAIWorkout } from './workoutSchema';
 
 export interface GenerationDelta {
   type: 'text' | 'usage' | 'complete' | 'error';
@@ -114,6 +115,9 @@ export function subscribe(
 export interface KickoffOpts {
   prisma: PrismaClient;
   userId: string;
+  /** "program" (multi-week) or "workout" (single day). Selects the
+   *  output parser and is persisted on the row. */
+  kind: 'program' | 'workout';
   templateId: string | null;
   templateName: string | null;
   userInput: string;
@@ -139,6 +143,7 @@ export async function kickoffGeneration(opts: KickoffOpts): Promise<string> {
   const generation = await opts.prisma.aIGeneration.create({
     data: {
       userId: opts.userId,
+      kind: opts.kind,
       templateId: opts.templateId,
       templateName: opts.templateName,
       userInput: opts.userInput,
@@ -248,10 +253,11 @@ async function runGeneration(generationId: string, opts: KickoffOpts) {
   let parsedJson: string | null = null;
   let parseErr: string | null = null;
   if (!providerError && raw) {
-    const r = parseAIProgram(raw);
+    const r =
+      opts.kind === 'workout' ? parseAIWorkout(raw) : parseAIProgram(raw);
     if (r.ok) {
       parsedOk = true;
-      parsedJson = JSON.stringify(r.program);
+      parsedJson = JSON.stringify('workout' in r ? r.workout : r.program);
     } else {
       parseErr = r.reason;
     }

proof-of-work/lib/ai/workoutDraft.ts

@@ -0,0 +1,78 @@
+import { isCardioExercise } from '@/lib/exerciseOptions';
+
+/**
+ * The ephemeral draft the "today's workout" flow hands to the New Workout
+ * form (via sessionStorage). One entry per exercise, with a working set
+ * count plus a single target weight/reps that we expand into N identical
+ * pre-filled sets. Shared by the producer (GenerateWorkoutClient) and the
+ * consumer (AiWorkoutPrefill) so the shape stays in sync.
+ */
+export interface AiWorkoutDraftExercise {
+  exerciseId: string;
+  sets: number;
+  reps?: number;
+  suggestedWeight?: number;
+  suggestedWeightUnit?: 'lbs' | 'kg';
+  rpe?: number;
+  gear?: number;
+  durationSeconds?: number;
+  notes?: string;
+}
+export interface AiWorkoutDraft {
+  name: string;
+  notes?: string;
+  exercises: AiWorkoutDraftExercise[];
+}
+
+export interface PrefillSet {
+  setNumber: number;
+  reps?: number;
+  weight?: number;
+  rpe?: number;
+  gear?: number;
+  durationSeconds?: number;
+  notes?: string;
+}
+export interface PrefillExercise<E> {
+  exercise: E;
+  sets: PrefillSet[];
+}
+
+/** Default working sets when the model omits a positive count. */
+const DEFAULT_SET_COUNT = 3;
+
+/**
+ * Expand a draft into pre-filled exercises against the user's library.
+ *
+ * - Exercises whose `exerciseId` isn't in the library are dropped (the
+ *   preview forces the user to map them first, so this is just defensive).
+ * - Each exercise becomes `sets` identical SetLogs seeded with the
+ *   suggested weight/reps.
+ * - Effort follows the app convention: cardio logs `gear` (1-5), every
+ *   other exercise logs `rpe`. We keep only the matching one so a stray
+ *   value on the wrong kind never reaches the form.
+ * - The coaching note rides only on the first set (avoids N copies).
+ */
+export function buildPrefillExercises<
+  E extends { id: string; type?: string | null; muscleGroups?: string | null },
+>(draft: AiWorkoutDraft, exercises: E[]): PrefillExercise<E>[] {
+  const byId = new Map(exercises.map((e) => [e.id, e]));
+  const out: PrefillExercise<E>[] = [];
+  for (const d of draft.exercises) {
+    const exercise = byId.get(d.exerciseId);
+    if (!exercise) continue;
+    const cardio = isCardioExercise(exercise);
+    const setCount = d.sets && d.sets > 0 ? d.sets : DEFAULT_SET_COUNT;
+    const sets: PrefillSet[] = Array.from({ length: setCount }, (_, i) => ({
+      setNumber: i + 1,
+      reps: d.reps,
+      weight: d.suggestedWeight,
+      rpe: cardio ? undefined : d.rpe,
+      gear: cardio ? d.gear : undefined,
+      durationSeconds: d.durationSeconds,
+      notes: i === 0 ? d.notes : undefined,
+    }));
+    out.push({ exercise, sets });
+  }
+  return out;
+}

proof-of-work/lib/ai/workoutPrompt.ts

@@ -0,0 +1,86 @@
+/**
+ * System-prompt builder for the "generate today's workout" flow. The
+ * sibling of systemPromptBase.ts (which targets multi-week programs).
+ *
+ * Job: force the single-workout JSON contract, ground suggested weights
+ * in the user's history, and respect the app's Gear-vs-RPE effort
+ * convention (cardio logs breathing Gear 1-5; everything else logs
+ * RPE 6-10).
+ */
+
+export interface WorkoutPromptOpts {
+  /** "lbs" | "kg" — default suggestedWeightUnit when the model omits one. */
+  weightUnit: 'lbs' | 'kg';
+  /** Whether the user's workout history is included in the prompt. */
+  hasHistoryContext: boolean;
+  /** True when the model is local (Ollama) — needs blunter, shorter rules. */
+  isLocalModel: boolean;
+  /** When refining, the prior suggestion's JSON. Present → revision mode. */
+  priorWorkoutJson?: string;
+}
+
+export function buildWorkoutSystemPrompt(opts: WorkoutPromptOpts): string {
+  const lines: string[] = [];
+
+  lines.push(
+    '# ROLE',
+    '',
+    "You are a strength & conditioning coach building ONE training session for today from the user's brain-dump. Turn their loose description into a concrete, ready-to-log workout.",
+    '',
+    '# OUTPUT CONTRACT (mandatory)',
+    '',
+    '1. Reply with EXACTLY ONE JSON object matching the OUTPUT SHAPE. No prose before or after. No ```json fences.',
+    '2. Every exercise must use an `exerciseId` from the LIBRARY block. NEVER invent ids. If nothing fits, pick the closest match and explain the substitution in `notes`.',
+    '3. Honor what the user asked for: include the exercises they named, with the set counts / emphasis they specified. Add sensible accessory work only if they asked you to fill out a body part (e.g. "let\'s do biceps and triceps").',
+    `4. Every resistance exercise MUST have a \`suggestedWeight\` (a number) and a target \`reps\`. Cardio, stretching, and bodyweight exercises set \`suggestedWeight\` to null.`,
+    `5. Express \`suggestedWeight\` in THAT exercise's \`unit\` from the LIBRARY block, and set \`suggestedWeightUnit\` to match it (default "${opts.weightUnit}" if none is shown). Don't convert — give the number in the exercise's own unit.`,
+    '6. `sets` is the number of working sets to pre-fill (e.g. the user\'s "4 working sets" → sets: 4).',
+    '7. EFFORT: for CARDIO exercises set `gear` (1-5 breathing gear) and leave `rpe` null. For everything else set `rpe` (1-10) and leave `gear` null.',
+    '8. Use `durationSeconds` instead of `reps` for timed work (holds, carries, intervals).',
+    '9. `notes` is for a short coaching cue — one sentence, optional.',
+    '10. Keep it to a single realistic session (typically 3-8 exercises). Do NOT invent multiple days or weeks — this is ONE workout.',
+  );
+
+  if (opts.hasHistoryContext) {
+    lines.push(
+      '',
+      '# USING THE HISTORY BLOCK',
+      '',
+      "The HISTORY block below summarizes the user's last 90 days. Use it to:",
+      '- Set `suggestedWeight` near their recent working weights for that exercise, NOT round numbers from nowhere.',
+      '- Nudge progressive overload where appropriate (small jump if a lift is moving; hold or deload if STAGNANT).',
+      '- Match the rep ranges and effort they tend to train at.',
+      "- If an exercise they named has no history, estimate conservatively and say so in `notes`.",
+    );
+  } else {
+    lines.push(
+      '',
+      '# WEIGHT GUIDANCE WITHOUT HISTORY',
+      '',
+      `Without prior data, set conservative \`suggestedWeight\` values (round gym increments; 5${opts.weightUnit} jumps, 2.5${opts.weightUnit} for small accessories) and add a coaching note like "adjust to leave 2-3 reps in reserve" so the user knows it's a starting estimate.`,
+    );
+  }
+
+  if (opts.priorWorkoutJson) {
+    lines.push(
+      '',
+      '# REVISION MODE',
+      '',
+      'The user already has the workout below and wants you to change it. Apply their requested change and re-emit the COMPLETE revised workout as one JSON object (not a diff). Keep everything they did not ask to change.',
+      '',
+      'CURRENT WORKOUT:',
+      opts.priorWorkoutJson,
+    );
+  }
+
+  if (opts.isLocalModel) {
+    lines.push(
+      '',
+      '# LOCAL MODEL REMINDER',
+      '',
+      'You are running locally with limited reasoning. Build the simplest valid single-session workout that matches the request. Do not overthink. JSON only.',
+    );
+  }
+
+  return lines.join('\n');
+}

proof-of-work/lib/ai/workoutSchema.ts

@@ -0,0 +1,124 @@
+import { z } from 'zod';
+import { extractJson } from './programSchema';
+
+/**
+ * The shape we ask LLMs to produce for a SINGLE day's workout (the
+ * "generate today's workout" flow). Distinct from the multi-week
+ * AIProgram in programSchema.ts.
+ *
+ * This does NOT map onto a DB table directly: the user reviews/edits the
+ * suggestion, then it pre-populates the normal New Workout form (nothing
+ * is persisted until they save through the regular workout path). So the
+ * shape is optimized for "pre-fill a logger" not "INSERT a Program".
+ *
+ * Per exercise we ask for a working `sets` count plus a single target
+ * `reps` / `suggestedWeight` — the hand-off expands that into N identical
+ * pre-filled SetLogs. (No warmup/ramping distinction in v1.)
+ *
+ * `exerciseId` is nullable: the model picks from the user's library when
+ * it can, but may suggest something not in the library (the preview
+ * prompts the user to map it). `exerciseName` is REQUIRED as the display
+ * label + fuzzy-match fallback.
+ */
+
+export const aiWorkoutExerciseSchema = z.object({
+  exerciseId: z.string().nullable(),
+  exerciseName: z.string().min(1),
+  order: z.number().int().nonnegative(),
+  /// Number of working sets to pre-fill. Defaults to 3 in the hand-off
+  /// if the model omits it.
+  sets: z.number().int().positive().optional().nullable(),
+  /// Target reps per set (the user overwrites with what they actually
+  /// did). Omit for time/distance-based work.
+  reps: z.number().int().positive().optional().nullable(),
+  /// Suggested working weight. Null for cardio / bodyweight / stretching.
+  suggestedWeight: z.number().nonnegative().optional().nullable(),
+  /// "lbs" | "kg". Optional; hand-off falls back to the user's
+  /// defaultWeightUnit when null.
+  suggestedWeightUnit: z.enum(['lbs', 'kg']).optional().nullable(),
+  /// Strength effort (1-10). The hand-off keeps this only for non-cardio
+  /// exercises (cardio uses `gear`).
+  rpe: z.number().int().min(1).max(10).optional().nullable(),
+  /// Cardio breathing gear (1-5). The hand-off keeps this only for
+  /// cardio exercises (strength uses `rpe`).
+  gear: z.number().int().min(1).max(5).optional().nullable(),
+  /// Target duration in seconds for time-based work (e.g. a hold).
+  durationSeconds: z.number().int().positive().optional().nullable(),
+  notes: z.string().optional().nullable(),
+});
+
+export const aiWorkoutSchema = z.object({
+  name: z.string().min(1),
+  notes: z.string().optional().nullable(),
+  exercises: z.array(aiWorkoutExerciseSchema),
+});
+
+export type AIWorkout = z.infer<typeof aiWorkoutSchema>;
+export type AIWorkoutExercise = z.infer<typeof aiWorkoutExerciseSchema>;
+
+/**
+ * JSON-schema-ish doc pasted into the system prompt so the model knows
+ * the exact shape to emit (same approach as PROGRAM_OUTPUT_SHAPE — not a
+ * provider "structured output" mode, since Ollama support is uneven).
+ */
+export const WORKOUT_OUTPUT_SHAPE = `{
+  "name": "<string, e.g. Upper Body — Shoulder Focus>",
+  "notes": "<string, optional, one-line session summary>",
+  "exercises": [
+    {
+      "exerciseId": "<string — REQUIRED — must be an id from the LIBRARY block. If no library exercise fits, pick the closest match and explain in notes; do NOT invent ids.>",
+      "exerciseName": "<string, the canonical name from the library>",
+      "order": <int >= 0>,
+      "sets": <int >= 1, number of working sets>,
+      "reps": <int, target reps per set; omit for time/distance work>,
+      "suggestedWeight": <number, working weight in the exercise's LIBRARY \`unit\`; omit/null for cardio, bodyweight, stretching>,
+      "suggestedWeightUnit": "<\\"lbs\\" | \\"kg\\", optional — match the exercise's \`unit\` from the LIBRARY>",
+      "rpe": <int 1-10, strength effort; use for NON-cardio exercises>,
+      "gear": <int 1-5, cardio breathing gear; use for CARDIO exercises instead of rpe>,
+      "durationSeconds": <int, optional, for timed holds/intervals>,
+      "notes": "<string, optional, short coaching cue>"
+    }
+  ]
+}`;
+
+/**
+ * Parse + validate a model's raw response into an AIWorkout. Returns a
+ * clean workout or a structured error. Mirrors parseAIProgram.
+ */
+export function parseAIWorkout(
+  raw: string,
+):
+  | { ok: true; workout: AIWorkout }
+  | { ok: false; reason: string; json?: string } {
+  const json = extractJson(raw);
+  if (!json) {
+    return {
+      ok: false,
+      reason: 'Could not find a JSON object in the response.',
+    };
+  }
+  let obj: unknown;
+  try {
+    obj = JSON.parse(json);
+  } catch (e) {
+    return {
+      ok: false,
+      reason: `JSON parse error: ${(e as Error).message}`,
+      json,
+    };
+  }
+  const result = aiWorkoutSchema.safeParse(obj);
+  if (!result.success) {
+    return {
+      ok: false,
+      reason:
+        'JSON did not match the expected shape: ' +
+        result.error.errors
+          .slice(0, 5)
+          .map((e) => `${e.path.join('.')}: ${e.message}`)
+          .join('; '),
+      json,
+    };
+  }
+  return { ok: true, workout: result.data };
+}

proof-of-work/lib/exerciseOptions.ts

@@ -131,3 +131,19 @@ export function deriveTrackingFieldOptions(exercises: Exercise[]): Option[] {
 export function displayLabel(value: string): string {
   return titleCaseToken(value);
 }
+
+/**
+ * Cardio exercises log breathing "Gear" (1-5) instead of RPE (6-10) as their
+ * effort field. An exercise counts as cardio if its equipment type is "cardio"
+ * or it carries the "cardio" muscle group (e.g. Assault Bike, type
+ * "assault bike", is tagged cardio).
+ */
+export function isCardioExercise(exercise: {
+  type?: string | null;
+  muscleGroups?: string | null;
+}): boolean {
+  if (normalizeValue(exercise.type || "") === "cardio") return true;
+  return parseJsonArray(exercise.muscleGroups ?? null).some(
+    (group) => normalizeValue(group) === "cardio"
+  );
+}

proof-of-work/prisma/schema.prisma

@@ -116,7 +116,8 @@ model SetLog {
   reps               Int?
   weight             Float?
   weightUnit         String    @default("lbs")
-  rpe                Int?      // Rate of Perceived Exertion (1-10)
+  rpe                Int?      // Rate of Perceived Exertion (1-10) — non-cardio effort
+  gear               Int?      // breathing "gear" (1-5, Brian MacKenzie) — cardio effort
   durationSeconds    Int?      // for timed exercises (assault bike, jump rope, planks)
   distance           Float?    // for distance-based exercises
   distanceUnit       String?   // "mi", "km", "m"
@@ -421,13 +422,20 @@ model AIGeneration {
   userInput          String
   systemPrompt       String
   userPrompt         String
+  /// What this generation produces: "program" (multi-week Program) or
+  /// "workout" (a single day's workout the user pre-fills the log from).
+  /// Drives which parser the runner uses and which UI consumes the row.
+  /// Defaults to "program" so legacy rows read correctly post-migration.
+  kind               String     @default("program")
   /// Streamed-so-far text. Updated periodically by the background
   /// generator so navigating-away clients can resume display via
   /// polling. Final value matches `rawResponse` once status flips
   /// to 'completed' or 'failed'.
   progressText       String?
   rawResponse        String?
-  parsedProgram      String?    // JSON.stringify of the parsed structure
+  /// JSON.stringify of the parsed structure. An AIProgram when
+  /// kind="program", an AIWorkout when kind="workout".
+  parsedProgram      String?
   provider           String
   model              String
   tokensIn           Int?

proof-of-work/tests/ai-workoutDraft.test.ts

@@ -0,0 +1,103 @@
+import { describe, it, expect } from 'vitest';
+import { buildPrefillExercises, type AiWorkoutDraft } from '@/lib/ai/workoutDraft';
+
+// Minimal library shape buildPrefillExercises needs (id + cardio inputs).
+const lib = [
+  { id: 'press', type: 'barbell', muscleGroups: '["shoulders"]' },
+  { id: 'bike', type: 'cardio', muscleGroups: '[]' },
+  // Tagged cardio via muscleGroups even though the equipment type isn't.
+  { id: 'boxjump', type: 'bodyweight', muscleGroups: '["legs","cardio"]' },
+];
+
+describe('buildPrefillExercises', () => {
+  it('expands a strength exercise into N sets with weight+reps and RPE only', () => {
+    const draft: AiWorkoutDraft = {
+      name: 'Push',
+      exercises: [
+        {
+          exerciseId: 'press',
+          sets: 4,
+          reps: 6,
+          suggestedWeight: 95,
+          rpe: 8,
+          gear: 3, // wrong-kind value — must be dropped for non-cardio
+        },
+      ],
+    };
+    const [ex] = buildPrefillExercises(draft, lib);
+    expect(ex.sets).toHaveLength(4);
+    expect(ex.sets.map((s) => s.setNumber)).toEqual([1, 2, 3, 4]);
+    for (const s of ex.sets) {
+      expect(s.weight).toBe(95);
+      expect(s.reps).toBe(6);
+      expect(s.rpe).toBe(8);
+      expect(s.gear).toBeUndefined();
+    }
+  });
+
+  it('uses gear (not rpe) for a cardio exercise', () => {
+    const draft: AiWorkoutDraft = {
+      name: 'Conditioning',
+      exercises: [
+        { exerciseId: 'bike', sets: 1, durationSeconds: 600, gear: 3, rpe: 8 },
+      ],
+    };
+    const [ex] = buildPrefillExercises(draft, lib);
+    expect(ex.sets[0].gear).toBe(3);
+    expect(ex.sets[0].rpe).toBeUndefined();
+    expect(ex.sets[0].durationSeconds).toBe(600);
+  });
+
+  it('treats an exercise tagged "cardio" in muscleGroups as cardio', () => {
+    const draft: AiWorkoutDraft = {
+      name: 'Plyo',
+      exercises: [{ exerciseId: 'boxjump', sets: 3, reps: 5, rpe: 7, gear: 2 }],
+    };
+    const [ex] = buildPrefillExercises(draft, lib);
+    expect(ex.sets[0].gear).toBe(2);
+    expect(ex.sets[0].rpe).toBeUndefined();
+  });
+
+  it('defaults to 3 sets when the count is missing or non-positive', () => {
+    const draft: AiWorkoutDraft = {
+      name: 'X',
+      exercises: [
+        { exerciseId: 'press', sets: 0, reps: 5, suggestedWeight: 100 },
+      ],
+    };
+    const [ex] = buildPrefillExercises(draft, lib);
+    expect(ex.sets).toHaveLength(3);
+  });
+
+  it('drops exercises whose id is not in the library', () => {
+    const draft: AiWorkoutDraft = {
+      name: 'X',
+      exercises: [
+        { exerciseId: 'ghost', sets: 3, reps: 5 },
+        { exerciseId: 'press', sets: 2, reps: 5, suggestedWeight: 100 },
+      ],
+    };
+    const out = buildPrefillExercises(draft, lib);
+    expect(out).toHaveLength(1);
+    expect(out[0].exercise.id).toBe('press');
+  });
+
+  it('puts the coaching note on the first set only', () => {
+    const draft: AiWorkoutDraft = {
+      name: 'X',
+      exercises: [
+        {
+          exerciseId: 'press',
+          sets: 3,
+          reps: 5,
+          suggestedWeight: 100,
+          notes: 'brace hard',
+        },
+      ],
+    };
+    const [ex] = buildPrefillExercises(draft, lib);
+    expect(ex.sets[0].notes).toBe('brace hard');
+    expect(ex.sets[1].notes).toBeUndefined();
+    expect(ex.sets[2].notes).toBeUndefined();
+  });
+});

proof-of-work/tests/ai-workoutSchema.test.ts

@@ -0,0 +1,88 @@
+import { describe, it, expect } from 'vitest';
+import { parseAIWorkout } from '@/lib/ai/workoutSchema';
+
+describe('parseAIWorkout', () => {
+  const valid = {
+    name: 'Upper — Shoulders',
+    notes: 'Overhead press focus',
+    exercises: [
+      {
+        exerciseId: 'cabc',
+        exerciseName: 'Overhead Press',
+        order: 0,
+        sets: 4,
+        reps: 6,
+        suggestedWeight: 95,
+        suggestedWeightUnit: 'lbs',
+        rpe: 8,
+      },
+      {
+        exerciseId: 'cdef',
+        exerciseName: 'Assault Bike',
+        order: 1,
+        sets: 1,
+        durationSeconds: 600,
+        gear: 3,
+      },
+    ],
+  };
+
+  it('accepts a valid single workout', () => {
+    const r = parseAIWorkout(JSON.stringify(valid));
+    expect(r.ok).toBe(true);
+    if (r.ok) {
+      expect(r.workout.name).toBe('Upper — Shoulders');
+      expect(r.workout.exercises).toHaveLength(2);
+      expect(r.workout.exercises[0].suggestedWeight).toBe(95);
+      expect(r.workout.exercises[1].gear).toBe(3);
+    }
+  });
+
+  it('accepts null exerciseId for unresolved exercises', () => {
+    const variant = structuredClone(valid);
+    variant.exercises[0].exerciseId = null as unknown as string;
+    const r = parseAIWorkout(JSON.stringify(variant));
+    expect(r.ok).toBe(true);
+  });
+
+  it('strips markdown fences and commentary', () => {
+    const wrapped =
+      "Here's today's session:\n\n```json\n" +
+      JSON.stringify(valid) +
+      '\n```\n\nEnjoy!';
+    const r = parseAIWorkout(wrapped);
+    expect(r.ok).toBe(true);
+  });
+
+  it('rejects when no JSON is present', () => {
+    const r = parseAIWorkout('the model just said hi');
+    expect(r.ok).toBe(false);
+    if (!r.ok) expect(r.reason).toMatch(/Could not find/);
+  });
+
+  it('rejects a parse-level syntax error inside balanced braces', () => {
+    const r = parseAIWorkout('{ "name": "x", }');
+    expect(r.ok).toBe(false);
+    if (!r.ok) expect(r.reason).toMatch(/parse error/i);
+  });
+
+  it('rejects when the shape is wrong (missing exercises)', () => {
+    const r = parseAIWorkout(JSON.stringify({ name: 'X' }));
+    expect(r.ok).toBe(false);
+    if (!r.ok) expect(r.reason).toMatch(/shape/);
+  });
+
+  it('rejects an out-of-range gear', () => {
+    const variant = structuredClone(valid);
+    variant.exercises[1].gear = 9; // gear is 1-5
+    const r = parseAIWorkout(JSON.stringify(variant));
+    expect(r.ok).toBe(false);
+  });
+
+  it('rejects an empty exercise name', () => {
+    const variant = structuredClone(valid);
+    variant.exercises[0].exerciseName = '';
+    const r = parseAIWorkout(JSON.stringify(variant));
+    expect(r.ok).toBe(false);
+  });
+});

proof-of-work/tests/cardio.test.ts

@@ -0,0 +1,28 @@
+import { describe, it, expect } from 'vitest';
+import { isCardioExercise } from '@/lib/exerciseOptions';
+
+describe('isCardioExercise', () => {
+  it('treats type "cardio" as cardio', () => {
+    expect(isCardioExercise({ type: 'cardio', muscleGroups: '["cardio"]' })).toBe(true);
+  });
+
+  it('treats the cardio muscle group as cardio even when type differs (Assault Bike)', () => {
+    expect(
+      isCardioExercise({ type: 'assault bike', muscleGroups: '["cardio","legs","back","shoulders"]' })
+    ).toBe(true);
+  });
+
+  it('is case/whitespace-insensitive on the muscle group', () => {
+    expect(isCardioExercise({ type: 'other', muscleGroups: '[" Cardio "]' })).toBe(true);
+  });
+
+  it('treats strength work (no cardio signal) as non-cardio', () => {
+    expect(isCardioExercise({ type: 'barbell', muscleGroups: '["back","biceps"]' })).toBe(false);
+  });
+
+  it('handles missing/empty fields without throwing', () => {
+    expect(isCardioExercise({})).toBe(false);
+    expect(isCardioExercise({ type: null, muscleGroups: null })).toBe(false);
+    expect(isCardioExercise({ type: '', muscleGroups: 'not json' })).toBe(false);
+  });
+});

proof-of-work/tests/routes-ai-workout.test.ts

@@ -0,0 +1,96 @@
+import { describe, it, expect, beforeEach, vi } from 'vitest';
+
+const { getCurrentUserMock } = vi.hoisted(() => ({
+  getCurrentUserMock: vi.fn(),
+}));
+vi.mock('@/lib/auth', async (orig) => {
+  const actual = (await orig()) as Record<string, unknown>;
+  return { ...actual, getCurrentUser: getCurrentUserMock };
+});
+
+import { NextRequest } from 'next/server';
+import { prisma } from '@/lib/prisma';
+import { POST } from '@/app/api/ai/generate-workout/route';
+
+const URL = 'http://x/api/ai/generate-workout';
+
+function jsonReq(body: unknown): NextRequest {
+  return new NextRequest(URL, {
+    method: 'POST',
+    headers: { 'content-type': 'application/json' },
+    body: JSON.stringify(body),
+  } as ConstructorParameters<typeof NextRequest>[1]);
+}
+function rawReq(rawBody: string): NextRequest {
+  return new NextRequest(URL, {
+    method: 'POST',
+    headers: { 'content-type': 'application/json' },
+    body: rawBody,
+  } as ConstructorParameters<typeof NextRequest>[1]);
+}
+
+async function makeUser(email: string) {
+  return prisma.user.create({
+    data: { email, passwordHash: 'fake', isAdmin: false },
+  });
+}
+
+beforeEach(async () => {
+  await prisma.aIGeneration.deleteMany();
+  await prisma.userPreferences.deleteMany();
+  await prisma.user.deleteMany();
+  getCurrentUserMock.mockReset();
+});
+
+describe('POST /api/ai/generate-workout — auth + validation', () => {
+  // These all return BEFORE the background runner is kicked off, so no
+  // real provider call happens. We deliberately don't exercise the 201
+  // path (it would spawn a detached generation).
+
+  it('401 when unauthenticated', async () => {
+    getCurrentUserMock.mockResolvedValue(null);
+    const res = await POST(jsonReq({ userInput: 'upper body' }));
+    expect(res.status).toBe(401);
+  });
+
+  it('400 on malformed JSON (not 500)', async () => {
+    const user = await makeUser('a@x');
+    getCurrentUserMock.mockResolvedValue(user);
+    const res = await POST(rawReq('{ not valid json'));
+    expect(res.status).toBe(400);
+  });
+
+  it('400 when userInput is missing', async () => {
+    const user = await makeUser('b@x');
+    getCurrentUserMock.mockResolvedValue(user);
+    const res = await POST(jsonReq({ includeHistory: true }));
+    expect(res.status).toBe(400);
+  });
+
+  it('400 when userInput is empty', async () => {
+    const user = await makeUser('c@x');
+    getCurrentUserMock.mockResolvedValue(user);
+    const res = await POST(jsonReq({ userInput: '' }));
+    expect(res.status).toBe(400);
+  });
+
+  it('400 with a malformed priorWorkout (fails the shared schema)', async () => {
+    const user = await makeUser('d@x');
+    getCurrentUserMock.mockResolvedValue(user);
+    // priorWorkout missing required `exercises` array → schema rejects.
+    const res = await POST(
+      jsonReq({ userInput: 'tweak it', priorWorkout: { name: 'X' } }),
+    );
+    expect(res.status).toBe(400);
+  });
+
+  it('400 when the user has no AI provider configured', async () => {
+    const user = await makeUser('e@x');
+    getCurrentUserMock.mockResolvedValue(user);
+    // Valid body, but no UserPreferences row → not configured.
+    const res = await POST(jsonReq({ userInput: 'upper body, shoulders' }));
+    expect(res.status).toBe(400);
+    const body = await res.json();
+    expect(body.error).toMatch(/not configured/i);
+  });
+});

proof-of-work/tests/routes-crud.test.ts

@@ -339,6 +339,45 @@ describe('POST /api/workouts', () => {
     expect(stored?.watts).toBe(157);
   });
 
+  it('persists gear (cardio breathing effort) on a set', async () => {
+    const alice = await makeUser({ email: 'a@x' });
+    const bike = await prisma.exercise.create({
+      data: {
+        userId: alice.id,
+        name: 'Assault Bike',
+        type: 'assault bike',
+        muscleGroups: '["cardio"]',
+        inputFields: '["sets","duration","distance","calories","watts","notes"]',
+      },
+    });
+    getCurrentUserMock.mockResolvedValue(alice);
+    const res = await postWorkout(
+      jsonReq('http://x/api/workouts', {
+        name: 'Conditioning',
+        sets: [{ exerciseId: bike.id, setNumber: 1, durationSeconds: 600, gear: 3 }],
+      }),
+    );
+    expect(res.status).toBe(201);
+    const body = await res.json();
+    expect(body.setLogs[0].gear).toBe(3);
+    const stored = await prisma.setLog.findFirst({ where: { workoutId: body.id } });
+    expect(stored?.gear).toBe(3);
+  });
+
+  it('rejects gear outside 1-5 via Zod with 400', async () => {
+    const alice = await makeUser({ email: 'a@x' });
+    const bike = await prisma.exercise.create({
+      data: { userId: alice.id, name: 'Assault Bike', type: 'assault bike', muscleGroups: '["cardio"]' },
+    });
+    getCurrentUserMock.mockResolvedValue(alice);
+    const res = await postWorkout(
+      jsonReq('http://x/api/workouts', {
+        sets: [{ exerciseId: bike.id, setNumber: 1, durationSeconds: 600, gear: 7 }],
+      }),
+    );
+    expect(res.status).toBe(400);
+  });
+
   it('rejects negative reps via Zod with 400', async () => {
     const alice = await makeUser({ email: 'a@x' });
     const bench = await prisma.exercise.create({
@@ -466,6 +505,28 @@ describe('PATCH /api/workouts/[id]', () => {
     const stored = await prisma.setLog.findFirst({ where: { workoutId: workout.id } });
     expect(stored?.watts).toBe(180);
   });
+
+  it('persists gear when replacing sets via PATCH', async () => {
+    const alice = await makeUser({ email: 'a@x' });
+    const bike = await prisma.exercise.create({
+      data: { userId: alice.id, name: 'Assault Bike', type: 'assault bike', muscleGroups: '["cardio"]' },
+    });
+    const workout = await prisma.workout.create({
+      data: { userId: alice.id, date: new Date(), name: 'Cond' },
+    });
+    getCurrentUserMock.mockResolvedValue(alice);
+    const res = await patchWorkout(
+      jsonReq(
+        'http://x/api/workouts/' + workout.id,
+        { sets: [{ exerciseId: bike.id, setNumber: 1, durationSeconds: 600, gear: 4 }] },
+        { method: 'PATCH' },
+      ),
+      { params: Promise.resolve({ id: workout.id }) },
+    );
+    expect(res.status).toBe(200);
+    const stored = await prisma.setLog.findFirst({ where: { workoutId: workout.id } });
+    expect(stored?.gear).toBe(4);
+  });
 });
 
 describe('POST /api/workouts/import/save', () => {

proof-of-work/types/index.ts

@@ -91,6 +91,7 @@ export type ParsedSet = {
   calories?: number | null;
   watts?: number | null;
   rpe?: number | null;
+  gear?: number | null;
   notes?: string | null;
 };
 

start9/0.4/docker_entrypoint.sh

@@ -77,6 +77,11 @@ if command -v sqlite3 >/dev/null 2>&1 && [ -f "$DB_PATH" ]; then
     sqlite3 "$DB_PATH" "ALTER TABLE SetLog ADD COLUMN watts INTEGER;"
   fi
 
+  if ! sqlite3 "$DB_PATH" "PRAGMA table_info('SetLog');" 2>/dev/null | grep -q "|gear|"; then
+    log "adding missing column SetLog.gear"
+    sqlite3 "$DB_PATH" "ALTER TABLE SetLog ADD COLUMN gear INTEGER;"
+  fi
+
   if ! sqlite3 "$DB_PATH" "PRAGMA table_info('Workout');" 2>/dev/null | grep -q "|deletedAt|"; then
     log "adding missing column Workout.deletedAt"
     sqlite3 "$DB_PATH" "ALTER TABLE Workout ADD COLUMN deletedAt DATETIME;"
@@ -216,6 +221,13 @@ if command -v sqlite3 >/dev/null 2>&1 && [ -f "$DB_PATH" ]; then
     sqlite3 "$DB_PATH" "ALTER TABLE AIGeneration ADD COLUMN durationMs INTEGER;"
   fi
 
+  # v1.2.0:6: single-workout generation. `kind` discriminates program vs
+  # workout rows; defaults to "program" so existing rows read correctly.
+  if ! sqlite3 "$DB_PATH" "PRAGMA table_info('AIGeneration');" 2>/dev/null | grep -q "|kind|"; then
+    log "adding AIGeneration.kind (default 'program')"
+    sqlite3 "$DB_PATH" "ALTER TABLE AIGeneration ADD COLUMN kind TEXT NOT NULL DEFAULT 'program';"
+  fi
+
   if ! sqlite3 "$DB_PATH" "PRAGMA table_info('ProgramExercise');" 2>/dev/null | grep -q "|suggestedWeight|"; then
     log "adding ProgramExercise.suggestedWeight + suggestedWeightUnit"
     sqlite3 "$DB_PATH" "ALTER TABLE ProgramExercise ADD COLUMN suggestedWeight REAL;"

start9/0.4/startos/versions/index.ts

@@ -19,6 +19,8 @@ import { v_1_2_0_1 } from './v1.2.0.1'
 import { v_1_2_0_2 } from './v1.2.0.2'
 import { v_1_2_0_3 } from './v1.2.0.3'
 import { v_1_2_0_4 } from './v1.2.0.4'
+import { v_1_2_0_5 } from './v1.2.0.5'
+import { v_1_2_0_6 } from './v1.2.0.6'
 
 /**
  * Version graph for the `proof-of-work` package.
@@ -76,9 +78,17 @@ import { v_1_2_0_4 } from './v1.2.0.4'
  *            column, added by the boot-time additive ALTER). Written through
  *            every set path; legacy watts in customMetrics stays readable and
  *            migrates on next save.
+ * v1.2.0:5 — Gear (breathing, 1-5, Brian MacKenzie) replaces RPE as the effort
+ *            field for cardio exercises (type "cardio" or "cardio" muscle
+ *            group); strength keeps RPE. New SetLog.gear column via boot ALTER.
+ * v1.2.0:6 — AI "generate today's workout": describe a single session and get
+ *            a ready-to-log workout (suggested weights/reps from history),
+ *            inline-edit + refine-with-AI, then pre-fill the workout log.
+ *            Reuses the generation spine via a new AIGeneration.kind
+ *            discriminant (boot ALTER, default "program"). No data changes.
  */
 export const versionGraph = VersionGraph.of({
-  current: v_1_2_0_4,
+  current: v_1_2_0_6,
   other: [
     v_1_0_0_1,
     v_1_0_0_2,
@@ -99,5 +109,7 @@ export const versionGraph = VersionGraph.of({
     v_1_2_0_1,
     v_1_2_0_2,
     v_1_2_0_3,
+    v_1_2_0_4,
+    v_1_2_0_5,
   ],
 })

start9/0.4/startos/versions/v1.2.0.5.ts

@@ -0,0 +1,25 @@
+import { IMPOSSIBLE, VersionInfo } from '@start9labs/start-sdk'
+
+/**
+ * v1.2.0:5 — Gear (breathing, 1-5) replaces RPE as the cardio effort field (2026-06-16).
+ *
+ * Cardio exercises now log a breathing "Gear" (1-5, per Brian MacKenzie)
+ * instead of RPE (6-10) as their effort field; non-cardio keeps RPE. An
+ * exercise counts as cardio if its equipment type is "cardio" or it carries
+ * the "cardio" muscle group (so Assault Bike, type "assault bike", qualifies).
+ *
+ * Additive schema change: the new nullable SetLog.gear column is added by the
+ * boot-time guarded ALTER in docker_entrypoint.sh (migration stays empty, like
+ * every other column add). Existing rpe data is untouched and still displays.
+ */
+export const v_1_2_0_5 = VersionInfo.of({
+  version: '1.2.0:5',
+  releaseNotes: {
+    en_US:
+      'Cardio exercises (assault bike, rower, ski erg, running, etc.) now log a breathing "Gear" (1-5) instead of RPE as their effort field. Strength exercises still use RPE. No data changes.',
+  },
+  migrations: {
+    up: async () => {},
+    down: IMPOSSIBLE,
+  },
+})

start9/0.4/startos/versions/v1.2.0.6.ts

@@ -0,0 +1,32 @@
+import { IMPOSSIBLE, VersionInfo } from '@start9labs/start-sdk'
+
+/**
+ * v1.2.0:6 — AI "generate today's workout" (2026-06-19).
+ *
+ * A new AI flow alongside program generation: describe a single session in
+ * plain words and get a ready-to-log workout back — exercises with suggested
+ * weights + target reps + set counts grounded in recent history. Inline-edit
+ * it, send a follow-up to refine it via the model, then "Use this workout" to
+ * pre-fill the normal New Workout form (nothing persists until you save).
+ *
+ * Reuses the existing generation spine (detached runner / SSE / lenient JSON /
+ * providers / history context) via a new AIGeneration.kind discriminant
+ * ("program" | "workout"). Single-workout rows are ephemeral and excluded from
+ * the program-shaped AI history.
+ *
+ * Additive schema change: the new AIGeneration.kind column (default "program")
+ * is added by the boot-time guarded ALTER in docker_entrypoint.sh, so this
+ * migration stays empty like every other column add. Existing rows read as
+ * "program"; no data changes.
+ */
+export const v_1_2_0_6 = VersionInfo.of({
+  version: '1.2.0:6',
+  releaseNotes: {
+    en_US:
+      'New: AI "Today\'s workout". Describe a session in plain words and get a ready-to-log workout with suggested weights and reps from your history. Edit it, refine it with the AI, then pre-fill the workout log. No data changes.',
+  },
+  migrations: {
+    up: async () => {},
+    down: IMPOSSIBLE,
+  },
+})