Previous value: -"The calendar date the workout happened, in the USER's timezone. Today only when the conversation is plainly about the session they just did. For anything logged after the fact, take the date from what the user said ('last Tuesday') and ask when that is unclear — never derive one from the server's clock. Everything time-based (weekly volume, the muscle load model, the consistency level) is anchored on it."New value: +"The calendar date the workout happened, in the USER's timezone: the day they did it, which for a session logged after the fact ('last Tuesday') is not today, and which the server's UTC clock does not know. Everything time-based (weekly volume, the muscle load model, the consistency level) is anchored on it."

Previous value: -"Catalog id of the exercise performed. Reuse an id the user already has (`exercise_catalog` in the coaching context, or list_exercises), otherwise a `slug` from search_exercise_pool verbatim. Minting a synonym for a movement that already has an id splits that movement's history in two."New value: +"Catalog id of the exercise performed: an id from the user's catalog (`exercise_catalog` in the coaching context, or list_exercises) or a `slug` from search_exercise_pool, character for character. A new id for a movement that already has one splits that movement's history in two."

Previous value: -"Metres covered, for a carry or an interval the user measured that way ('farmer's walk, 30 metres'). Use it instead of duration_sec when what they said was a distance, not a time; fill both only when they gave both."New value: +"Metres covered, for a carry or an interval the user measured that way ('farmer's walk, 30 metres'). Both distance_m and duration_sec are present only when both were measured."

Previous value: -"Seconds under load for a hold or carry (plank, dead hang, farmer's walk), or the duration of a timed interval. Isometric work is dosed in seconds, not reps. Use it when the user gave a time; when they gave a distance instead, use distance_m."New value: +"Seconds under load for a hold or carry (plank, dead hang, farmer's walk), or the duration of a timed interval. Isometric work is dosed in seconds, not reps. A distance-measured carry or interval is distance_m instead."

Previous value: -"True when the set was done one side at a time. Log it as ONE set whose `reps` holds ONE side's count — never doubled, never split into two sets — and set this flag so the number is read the way it was meant. A timed one-sided hold or carry is per-side too: fill duration_sec, leave reps empty. Logging the same exercise both ways stops it comparing to itself."New value: +"True when the set was done one side at a time. Such a set is ONE record whose `reps` (or duration_sec, for a one-sided hold or carry) is ONE side's count, neither doubled nor split into two sets. An exercise recorded both ways does not compare with itself."

Previous value: -"Reps actually completed. On a per-side set give ONE side's count and set is_per_side — never the doubled total, never two sets. Leave empty for a timed hold or carry and use duration_sec instead."New value: +"Reps actually completed. On a per-side set (is_per_side) this is ONE side's count: not the doubled total, and not split into two sets. Empty for a timed hold or carry, whose dose is duration_sec."

Previous value: -"Reps left in the tank at the end of the set; 0 means taken to failure. Fill it only when the user says how close to failure they got ('had two left', 'went to failure'); a bare 'that was hard' or 'felt easy' is not a number — leave it empty rather than guessing one. It is what tells the model the set's load fades slower."New value: +"Reps left in the tank at the end of the set as the user stated it ('had two left', 'went to failure' = 0); a bare 'that was hard' or 'felt easy' states no number, and then the field is empty. It is what tells the model the set's load fades slower."

Previous value: -"Perceived effort 0-10 — the same question as rir on the other scale (RPE 10 = RIR 0). Use it when the user talks in RPE; rir wins when both are given."New value: +"Perceived effort 0-10 — the same question as rir on the other scale (RPE 10 = RIR 0). It holds a value the user gave in RPE; when both are present, rir takes precedence."

Previous value: -"The client's own stable id for this set, if it has one. Only the AIm app sends this: it lets a later correction be traced to the set that moved, because set_number is renumbered when a set is removed. Other clients have none and leave it out."New value: +"The client's own stable id for this set, if it has one. Only the AIm app sends this: it lets a later correction be traced to the set that moved, because set_number is renumbered when a set is removed. Absent from other clients."

Previous value: -"`working` for real sets. Mark ramp-up sets `warmup`: they are excluded from volume, weekly set counts and the muscle load model, so logging them as working inflates every number the coach reads back. A drop set is not one set: log each weight stage as its own entry in the order performed, the lighter stages `dropset`."New value: +"`working` for real sets, `warmup` for ramp-up sets: warm-ups are excluded from volume, weekly set counts and the muscle load model, and counted as working they inflate every number read back. A drop set is several records, one per weight stage in the order performed, the lighter stages typed `dropset`."

Previous value: -"Load in kilograms: the whole bar for barbell work, ONE dumbbell or kettlebell for dumbbell work. Omit for unloaded bodyweight sets; give only the added weight when the user hangs a plate off a belt."New value: +"Load in kilograms: the whole bar for barbell work, ONE dumbbell or kettlebell for dumbbell work. Absent for unloaded bodyweight sets; for weighted bodyweight work (a plate on a belt) it is the added weight only."

Previous value: -"When the session started, with timezone offset, if the user actually said. Omit it rather than inventing a plausible time."New value: +"When the session started, with timezone offset, as the user stated it; absent when they did not say."

Previous value: -"The client's own stable id for this set, if it has one. Only the AIm app sends this: it lets a later correction be traced to the set that moved, because set_number is renumbered when a set is removed. Never invent one — omit it."New value: +"The client's own stable id for this set, if it has one. Only the AIm app sends this: it lets a later correction be traced to the set that moved, because set_number is renumbered when a set is removed. Other clients have none and leave it out."

Added value: +{
+  "anyOf": [
+    {
+      "maxLength": 128,
+      "type": "string"
+    },
+    {
+      "type": "null"
+    }
+  ],
+  "default": null,
+  "description": "The client's own stable id for this set, if it has one. Only the AIm app sends this: it lets a later correction be traced to the set that moved, because set_number is renumbered when a set is removed. Never invent one — omit it."
+}