feat(kv): set_if compare-and-swap for KV (per-app + shared + SDK) (Track A M5)

No atomic CAS existed, so concurrent writers to the same KV key raced. Add
set_if(key, expected, new): writes only when the current value equals expected,
or (expected absent) only when the key is missing — the primitive for lock-free
counters / optimistic concurrency.

- KvService + GroupKvService traits gain set_if with a defaulted "unsupported"
  impl (Noop + other impls untouched); the real services override it.
- kv_repo/group_kv_repo: atomic set_if — a single UPDATE ... WHERE value =
  $expected (JSONB semantic equality) or INSERT ... ON CONFLICT DO NOTHING.
  Atomic without a transaction; returns whether the write happened.
- Services enforce the same value-size cap + (group) writes-authed + row/byte
  quotas as set; the event fires only on a successful swap.
- Executor Rhai SDK: kv::collection(c).set_if(key, expected, new) and the
  shared_collection variant; Rhai () for expected means "only if absent".

Pinned by kv_service/group_kv_service set_if unit tests (CAS + fail-closed) +
sdk_kv::kv_set_if_compare_and_swap (through a real script). No migration.
docs/sdk-shape.md documents the primitive.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

crates/executor-core/src/sdk/kv.rs

@@ -5,6 +5,8 @@
 //! widgets.set("k", #{ n: 1 });
 //! let v = widgets.get("k");      // value or () if absent
 //! if widgets.has("k") { ... }
+//! widgets.set_if("k", (), #{ n: 0 });   // insert only if ABSENT -> bool
+//! widgets.set_if("k", old, new);        // swap only if current == old -> bool
 //! widgets.delete("k");            // bool (was-present)
 //! let page = widgets.list();      // returns #{ keys: [...], next_cursor: () }
 //! ```
@@ -105,6 +107,7 @@ pub(super) fn register(engine: &mut RhaiEngine, services: &Services, cx: Arc<Sdk
 
     register_get(engine);
     register_set(engine);
+    register_set_if(engine);
     register_has(engine);
     register_delete(engine);
     register_list(engine);
@@ -114,11 +117,23 @@ pub(super) fn register(engine: &mut RhaiEngine, services: &Services, cx: Arc<Sdk
 
     register_group_get(engine);
     register_group_set(engine);
+    register_group_set_if(engine);
     register_group_has(engine);
     register_group_delete(engine);
     register_group_list(engine);
 }
 
+/// Map a Rhai `expected` argument to the CAS precondition: Rhai unit `()` means
+/// "expected ABSENT" (insert-if-absent); any other value is the expected current
+/// value.
+fn expected_from_dynamic(expected: &Dynamic) -> Option<serde_json::Value> {
+    if expected.is_unit() {
+        None
+    } else {
+        Some(dynamic_to_json(expected))
+    }
+}
+
 fn register_get(engine: &mut RhaiEngine) {
     engine.register_fn(
         "get",
@@ -145,6 +160,27 @@ fn register_set(engine: &mut RhaiEngine) {
     );
 }
 
+fn register_set_if(engine: &mut RhaiEngine) {
+    // `handle.set_if(key, expected, new)` — CAS. `expected = ()` means "only if
+    // absent". Returns `true` if the swap happened, `false` if the precondition
+    // failed.
+    engine.register_fn(
+        "set_if",
+        |handle: &mut KvHandle,
+         key: &str,
+         expected: Dynamic,
+         new: Dynamic|
+         -> Result<bool, Box<EvalAltResult>> {
+            let h = handle.clone();
+            let exp = expected_from_dynamic(&expected);
+            let new = dynamic_to_json(&new);
+            block_on("kv", async move {
+                h.service.set_if(&h.cx, &h.collection, key, exp, new).await
+            })
+        },
+    );
+}
+
 fn register_has(engine: &mut RhaiEngine) {
     engine.register_fn(
         "has",
@@ -243,6 +279,24 @@ fn register_group_set(engine: &mut RhaiEngine) {
     );
 }
 
+fn register_group_set_if(engine: &mut RhaiEngine) {
+    engine.register_fn(
+        "set_if",
+        |handle: &mut GroupKvHandle,
+         key: &str,
+         expected: Dynamic,
+         new: Dynamic|
+         -> Result<bool, Box<EvalAltResult>> {
+            let h = handle.clone();
+            let exp = expected_from_dynamic(&expected);
+            let new = dynamic_to_json(&new);
+            block_on("kv", async move {
+                h.service.set_if(&h.cx, &h.collection, key, exp, new).await
+            })
+        },
+    );
+}
+
 fn register_group_has(engine: &mut RhaiEngine) {
     engine.register_fn(
         "has",

crates/executor-core/tests/sdk_kv.rs

@@ -52,6 +52,27 @@ impl KvService for InMemoryKv {
         Ok(())
     }
 
+    async fn set_if(
+        &self,
+        cx: &SdkCallCx,
+        collection: &str,
+        key: &str,
+        expected: Option<Value>,
+        new: Value,
+    ) -> Result<bool, KvError> {
+        let mut data = self.data.lock().await;
+        let k = (cx.app_id, collection.to_string(), key.to_string());
+        let matches = match (&expected, data.get(&k)) {
+            (None, None) => true,
+            (Some(exp), Some(cur)) => exp == cur,
+            _ => false,
+        };
+        if matches {
+            data.insert(k, new);
+        }
+        Ok(matches)
+    }
+
     async fn delete(&self, cx: &SdkCallCx, collection: &str, key: &str) -> Result<bool, KvError> {
         Ok(self
             .data
@@ -167,6 +188,27 @@ async fn kv_set_then_get_round_trip() {
     assert_eq!(body, json!({ "n": 1 }));
 }
 
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn kv_set_if_compare_and_swap() {
+    let engine = make_engine();
+    let app = AppId::new();
+    // `set_if(key, (), new)` inserts only when absent; `set_if(key, expected,
+    // new)` swaps only on match. Returns a bool each time.
+    let src = r#"
+        let c = kv::collection("counters");
+        let first = c.set_if("n", (), 1);      // absent -> inserts, true
+        let dup   = c.set_if("n", (), 2);      // present -> false
+        let bad   = c.set_if("n", 99, 3);      // mismatch -> false
+        let good  = c.set_if("n", 1, 3);       // match   -> swaps, true
+        #{ first: first, dup: dup, bad: bad, good: good, final: c.get("n") }
+    "#;
+    let body = run_script(engine, src, baseline_request(app)).await;
+    assert_eq!(
+        body,
+        json!({ "first": true, "dup": false, "bad": false, "good": true, "final": 3 })
+    );
+}
+
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn kv_get_missing_returns_unit() {
     let engine = make_engine();

crates/manager-core/src/group_kv_repo.rs

@@ -39,6 +39,18 @@ pub trait GroupKvRepo: Send + Sync {
         value: serde_json::Value,
     ) -> Result<Option<serde_json::Value>, GroupKvRepoError>;
 
+    /// Compare-and-swap. Atomically writes `new` iff the current value equals
+    /// `expected` (`expected = None` → iff the key is ABSENT). Returns `true`
+    /// when the write happened. One SQL statement — atomic without a tx.
+    async fn set_if(
+        &self,
+        group_id: GroupId,
+        collection: &str,
+        key: &str,
+        expected: Option<serde_json::Value>,
+        new: serde_json::Value,
+    ) -> Result<bool, GroupKvRepoError>;
+
     /// Returns the deleted value if present, `None` if the row didn't exist.
     async fn delete(
         &self,
@@ -143,6 +155,43 @@ impl GroupKvRepo for PostgresGroupKvRepo {
         Ok(row.and_then(|(v,)| v))
     }
 
+    async fn set_if(
+        &self,
+        group_id: GroupId,
+        collection: &str,
+        key: &str,
+        expected: Option<serde_json::Value>,
+        new: serde_json::Value,
+    ) -> Result<bool, GroupKvRepoError> {
+        let affected = match expected {
+            Some(exp) => sqlx::query(
+                "UPDATE group_kv_entries SET value = $4, updated_at = NOW() \
+                     WHERE group_id = $1 AND collection = $2 AND key = $3 AND value = $5",
+            )
+            .bind(group_id.into_inner())
+            .bind(collection)
+            .bind(key)
+            .bind(new)
+            .bind(exp)
+            .execute(&self.pool)
+            .await?
+            .rows_affected(),
+            None => sqlx::query(
+                "INSERT INTO group_kv_entries (group_id, collection, key, value) \
+                     VALUES ($1, $2, $3, $4) \
+                     ON CONFLICT (group_id, collection, key) DO NOTHING",
+            )
+            .bind(group_id.into_inner())
+            .bind(collection)
+            .bind(key)
+            .bind(new)
+            .execute(&self.pool)
+            .await?
+            .rows_affected(),
+        };
+        Ok(affected == 1)
+    }
+
     async fn delete(
         &self,
         group_id: GroupId,

crates/manager-core/src/group_kv_service.rs

@@ -263,6 +263,66 @@ impl GroupKvService for GroupKvServiceImpl {
         Ok(())
     }
 
+    async fn set_if(
+        &self,
+        cx: &SdkCallCx,
+        collection: &str,
+        key: &str,
+        expected: Option<serde_json::Value>,
+        new: serde_json::Value,
+    ) -> Result<bool, GroupKvError> {
+        let group_id = self.owning_group(cx, collection).await?;
+        let encoded_len = serde_json::to_vec(&new)
+            .map(|v| v.len())
+            .map_err(|e| GroupKvError::Backend(format!("encode value: {e}")))?;
+        if encoded_len > self.max_value_bytes {
+            return Err(GroupKvError::ValueTooLarge {
+                limit: self.max_value_bytes,
+                actual: encoded_len,
+            });
+        }
+        self.check_write(cx, group_id).await?;
+        // Same quota rules as `set` (best-effort pre-check; the swap itself is
+        // atomic in the repo). A swap that inserts a NEW key (expected absent)
+        // must fit the row ceiling; the byte ceiling uses the projected total.
+        let existing = self.repo.get(group_id, collection, key).await?;
+        if expected.is_none() && existing.is_none() {
+            let count = self.repo.count_rows(group_id).await?;
+            if count >= self.max_rows {
+                return Err(GroupKvError::QuotaExceeded {
+                    limit: usize::try_from(self.max_rows).unwrap_or(usize::MAX),
+                    actual: usize::try_from(count).unwrap_or(usize::MAX),
+                });
+            }
+        }
+        let old_len = existing
+            .as_ref()
+            .and_then(|v| serde_json::to_vec(v).ok())
+            .map_or(0u64, |b| b.len() as u64);
+        let used = self.repo.total_bytes(group_id).await?;
+        let projected = used.saturating_sub(old_len) + encoded_len as u64;
+        if projected > self.max_total_bytes {
+            return Err(GroupKvError::TotalBytesQuotaExceeded {
+                limit: self.max_total_bytes,
+                actual: projected,
+            });
+        }
+        let op = if expected.is_some() {
+            "update"
+        } else {
+            "insert"
+        };
+        let swapped = self
+            .repo
+            .set_if(group_id, collection, key, expected, new.clone())
+            .await?;
+        if swapped {
+            self.emit_shared(cx, group_id, op, collection, key, Some(new))
+                .await;
+        }
+        Ok(swapped)
+    }
+
     async fn delete(
         &self,
         cx: &SdkCallCx,
@@ -348,6 +408,27 @@ mod tests {
                 .insert((group_id, collection.to_string(), key.to_string()), value))
         }
 
+        async fn set_if(
+            &self,
+            group_id: GroupId,
+            collection: &str,
+            key: &str,
+            expected: Option<serde_json::Value>,
+            new: serde_json::Value,
+        ) -> Result<bool, GroupKvRepoError> {
+            let mut data = self.data.lock().await;
+            let k = (group_id, collection.to_string(), key.to_string());
+            let matches = match (&expected, data.get(&k)) {
+                (None, None) => true,
+                (Some(exp), Some(cur)) => exp == cur,
+                _ => false,
+            };
+            if matches {
+                data.insert(k, new);
+            }
+            Ok(matches)
+        }
+
         async fn delete(
             &self,
             group_id: GroupId,
@@ -643,4 +724,56 @@ mod tests {
         let err = kv.get(&cx, "", "k").await.unwrap_err();
         assert!(matches!(err, GroupKvError::InvalidCollection));
     }
+
+    #[tokio::test]
+    async fn set_if_is_compare_and_swap_and_writes_fail_closed() {
+        let app = AppId::new();
+        let group = GroupId::new();
+        let mut resolver = FakeResolver::default();
+        resolver.map.insert((app, "catalog".into()), group);
+        let kv = svc(resolver);
+        let cx = cx_with(app, Some(owner()));
+
+        // Insert-if-absent, then swap-if-equal (as the per-app path).
+        assert!(kv
+            .set_if(&cx, "catalog", "k", None, serde_json::json!(1))
+            .await
+            .unwrap());
+        assert!(!kv
+            .set_if(&cx, "catalog", "k", None, serde_json::json!(2))
+            .await
+            .unwrap());
+        assert!(!kv
+            .set_if(
+                &cx,
+                "catalog",
+                "k",
+                Some(serde_json::json!(9)),
+                serde_json::json!(3)
+            )
+            .await
+            .unwrap());
+        assert!(kv
+            .set_if(
+                &cx,
+                "catalog",
+                "k",
+                Some(serde_json::json!(1)),
+                serde_json::json!(3)
+            )
+            .await
+            .unwrap());
+        assert_eq!(
+            kv.get(&cx, "catalog", "k").await.unwrap(),
+            Some(serde_json::json!(3))
+        );
+
+        // A shared-collection CAS still fails closed for an anonymous writer.
+        let anon = cx_with(app, None);
+        let err = kv
+            .set_if(&anon, "catalog", "k", None, serde_json::json!(4))
+            .await
+            .unwrap_err();
+        assert!(matches!(err, GroupKvError::Forbidden));
+    }
 }

crates/manager-core/src/kv_repo.rs

@@ -39,6 +39,19 @@ pub trait KvRepo: Send + Sync {
         value: serde_json::Value,
     ) -> Result<Option<serde_json::Value>, KvRepoError>;
 
+    /// Compare-and-swap. Atomically writes `new` iff the current value equals
+    /// `expected` (`expected = None` → iff the key is ABSENT). Returns `true`
+    /// when the write happened. A single SQL statement, so it is atomic without a
+    /// transaction.
+    async fn set_if(
+        &self,
+        app_id: AppId,
+        collection: &str,
+        key: &str,
+        expected: Option<serde_json::Value>,
+        new: serde_json::Value,
+    ) -> Result<bool, KvRepoError>;
+
     /// Returns the deleted value if present, `None` if the row didn't
     /// exist. The caller turns the `bool was-present` part into the
     /// SDK's return value; the `Option<value>` part feeds the
@@ -131,6 +144,45 @@ impl KvRepo for PostgresKvRepo {
         Ok(row.and_then(|(v,)| v))
     }
 
+    async fn set_if(
+        &self,
+        app_id: AppId,
+        collection: &str,
+        key: &str,
+        expected: Option<serde_json::Value>,
+        new: serde_json::Value,
+    ) -> Result<bool, KvRepoError> {
+        let affected = match expected {
+            // Swap iff the current value matches (JSONB `=` is semantic equality).
+            Some(exp) => sqlx::query(
+                "UPDATE kv_entries SET value = $4, updated_at = NOW() \
+                     WHERE app_id = $1 AND collection = $2 AND key = $3 AND value = $5",
+            )
+            .bind(app_id.into_inner())
+            .bind(collection)
+            .bind(key)
+            .bind(new)
+            .bind(exp)
+            .execute(&self.pool)
+            .await?
+            .rows_affected(),
+            // Insert iff absent.
+            None => sqlx::query(
+                "INSERT INTO kv_entries (app_id, collection, key, value) \
+                     VALUES ($1, $2, $3, $4) \
+                     ON CONFLICT (app_id, collection, key) DO NOTHING",
+            )
+            .bind(app_id.into_inner())
+            .bind(collection)
+            .bind(key)
+            .bind(new)
+            .execute(&self.pool)
+            .await?
+            .rows_affected(),
+        };
+        Ok(affected == 1)
+    }
+
     async fn delete(
         &self,
         app_id: AppId,

crates/manager-core/src/kv_service.rs

@@ -186,6 +186,59 @@ impl KvService for KvServiceImpl {
         Ok(())
     }
 
+    async fn set_if(
+        &self,
+        cx: &SdkCallCx,
+        collection: &str,
+        key: &str,
+        expected: Option<serde_json::Value>,
+        new: serde_json::Value,
+    ) -> Result<bool, KvError> {
+        validate_collection(collection)?;
+        let encoded_len = serde_json::to_vec(&new)
+            .map(|v| v.len())
+            .map_err(|e| KvError::Backend(format!("encode value: {e}")))?;
+        if encoded_len > self.max_value_bytes {
+            return Err(KvError::ValueTooLarge {
+                limit: self.max_value_bytes,
+                actual: encoded_len,
+            });
+        }
+        self.check_write(cx).await?;
+        // `insert` when the precondition was "absent", else `update`. The op is
+        // only emitted on a successful swap.
+        let op = if expected.is_some() {
+            "update"
+        } else {
+            "insert"
+        };
+        let swapped = self
+            .repo
+            .set_if(cx.app_id, collection, key, expected, new.clone())
+            .await?;
+        if swapped {
+            // Non-transactional with the write (same caveat as `set`).
+            if let Err(e) = self
+                .events
+                .emit(
+                    cx,
+                    ServiceEvent {
+                        source: "kv",
+                        op,
+                        collection: Some(collection.to_string()),
+                        key: Some(key.to_string()),
+                        payload: Some(new),
+                        old_payload: None,
+                    },
+                )
+                .await
+            {
+                tracing::error!(error = %e, source = "kv", op, event_emit_failure = true, "event emit failed");
+            }
+        }
+        Ok(swapped)
+    }
+
     async fn delete(&self, cx: &SdkCallCx, collection: &str, key: &str) -> Result<bool, KvError> {
         validate_collection(collection)?;
         self.check_write(cx).await?;
@@ -283,6 +336,28 @@ mod tests {
                 .insert((app_id, collection.to_string(), key.to_string()), value))
         }
 
+        async fn set_if(
+            &self,
+            app_id: AppId,
+            collection: &str,
+            key: &str,
+            expected: Option<serde_json::Value>,
+            new: serde_json::Value,
+        ) -> Result<bool, KvRepoError> {
+            // Holding the map lock makes the compare-and-set atomic in-memory.
+            let mut data = self.data.lock().await;
+            let k = (app_id, collection.to_string(), key.to_string());
+            let matches = match (&expected, data.get(&k)) {
+                (None, None) => true,                 // insert-if-absent
+                (Some(exp), Some(cur)) => exp == cur, // swap-if-equal
+                _ => false,
+            };
+            if matches {
+                data.insert(k, new);
+            }
+            Ok(matches)
+        }
+
         async fn delete(
             &self,
             app_id: AppId,
@@ -464,6 +539,63 @@ mod tests {
         assert!(matches!(err, KvError::InvalidCollection));
     }
 
+    #[tokio::test]
+    async fn set_if_is_compare_and_swap() {
+        let kv = svc();
+        let cx = anon_cx(AppId::new());
+        // Insert-if-absent: expected () swaps only when the key is missing.
+        assert!(
+            kv.set_if(&cx, "cas", "k", None, serde_json::json!(1))
+                .await
+                .unwrap(),
+            "insert-if-absent must succeed on a missing key"
+        );
+        assert!(
+            !kv.set_if(&cx, "cas", "k", None, serde_json::json!(2))
+                .await
+                .unwrap(),
+            "insert-if-absent must fail once the key exists"
+        );
+        assert_eq!(
+            kv.get(&cx, "cas", "k").await.unwrap(),
+            Some(serde_json::json!(1))
+        );
+
+        // Swap-if-equal: wrong expected → no-op; correct expected → swaps.
+        assert!(
+            !kv.set_if(
+                &cx,
+                "cas",
+                "k",
+                Some(serde_json::json!(99)),
+                serde_json::json!(3)
+            )
+            .await
+            .unwrap(),
+            "a mismatched expected must not swap"
+        );
+        assert_eq!(
+            kv.get(&cx, "cas", "k").await.unwrap(),
+            Some(serde_json::json!(1))
+        );
+        assert!(
+            kv.set_if(
+                &cx,
+                "cas",
+                "k",
+                Some(serde_json::json!(1)),
+                serde_json::json!(3)
+            )
+            .await
+            .unwrap(),
+            "a matching expected must swap"
+        );
+        assert_eq!(
+            kv.get(&cx, "cas", "k").await.unwrap(),
+            Some(serde_json::json!(3))
+        );
+    }
+
     /// Load-bearing: a script with `cx.app_id = A` must NOT see
     /// entries inserted under `cx.app_id = B`. This is the cross-app
     /// isolation boundary; getting this wrong is a security

crates/shared/src/group_kv.rs

@@ -37,6 +37,25 @@ pub trait GroupKvService: Send + Sync {
         value: serde_json::Value,
     ) -> Result<(), GroupKvError>;
 
+    /// Compare-and-swap on a shared-collection key. Atomically writes `new` only
+    /// if the current value equals `expected` — or, when `expected` is `None`,
+    /// only if the key is ABSENT. Returns `true` on swap, `false` if the
+    /// precondition failed. Same reads-open / writes-authed + quota rules as
+    /// [`Self::set`]. Default: unsupported.
+    async fn set_if(
+        &self,
+        cx: &SdkCallCx,
+        collection: &str,
+        key: &str,
+        expected: Option<serde_json::Value>,
+        new: serde_json::Value,
+    ) -> Result<bool, GroupKvError> {
+        let _ = (cx, collection, key, expected, new);
+        Err(GroupKvError::Backend(
+            "set_if is not supported by this group-kv implementation".into(),
+        ))
+    }
+
     async fn delete(
         &self,
         cx: &SdkCallCx,

crates/shared/src/kv.rs

@@ -36,6 +36,25 @@ pub trait KvService: Send + Sync {
         value: serde_json::Value,
     ) -> Result<(), KvError>;
 
+    /// Compare-and-swap. Atomically writes `new` only if the current value
+    /// equals `expected` — or, when `expected` is `None`, only if the key is
+    /// ABSENT (insert-if-absent). Returns `true` if the swap happened, `false`
+    /// if the precondition failed (value differed / key already present).
+    /// Default: unsupported, so only storage-backed impls opt in.
+    async fn set_if(
+        &self,
+        cx: &SdkCallCx,
+        collection: &str,
+        key: &str,
+        expected: Option<serde_json::Value>,
+        new: serde_json::Value,
+    ) -> Result<bool, KvError> {
+        let _ = (cx, collection, key, expected, new);
+        Err(KvError::Backend(
+            "set_if is not supported by this kv implementation".into(),
+        ))
+    }
+
     async fn delete(&self, cx: &SdkCallCx, collection: &str, key: &str) -> Result<bool, KvError>;
 
     async fn has(&self, cx: &SdkCallCx, collection: &str, key: &str) -> Result<bool, KvError>;

docs/sdk-shape.md

@@ -57,8 +57,18 @@ returned by an `::collection(name)` constructor:
 let widgets = kv::collection("widgets");
 widgets.set("k", "v");
 let v = widgets.get("k");
+// Compare-and-swap (v1.2, Track A M5): atomic, returns a bool.
+widgets.set_if("k", (), "v0");   // insert only if ABSENT
+widgets.set_if("k", "v0", "v1"); // swap only if current == "v0"
 ```
 
+`set_if(key, expected, new)` is an atomic compare-and-swap: an `expected` of
+`()` means "only if the key is absent" (insert-if-absent); any other `expected`
+swaps only when the stored value equals it. Returns `true` if the write
+happened, `false` if the precondition failed — the primitive for lock-free
+counters / optimistic concurrency. Available on both `kv::collection` and
+`kv::shared_collection`.
+
 Not `kv::set("widgets", "k", "v")`. The handle is a Rhai custom type
 the service registers; method calls bind to that type. This: