Merge branch 'main' into repo-assist/perf-mapiasync-2026-03-23-18b8b82a487eb386

Author: dsyme

.github/aw/actions-lock.json

@@ -15,10 +15,15 @@
       "version": "v0.61.1",
       "sha": "be0029bbbaeef8c6bea6825f31d9593219b2dc28"
     },
-    "github/gh-aw/actions/setup@v0.61.2": {
+    "github/gh-aw-actions/setup@v0.62.5": {
+      "repo": "github/gh-aw-actions/setup",
+      "version": "v0.62.5",
+      "sha": "dc50be57c94373431b49d3d0927f318ac2bb5c4c"
+    },
+    "github/gh-aw/actions/setup@v0.62.5": {
       "repo": "github/gh-aw/actions/setup",
-      "version": "v0.61.2",
-      "sha": "d6f6273a03402cd530be35455a7823494b846d66"
+      "version": "v0.62.5",
+      "sha": "48d8fdfddc8cad854ac0c70ceb573f09fb8f9c9b"
     }
   }
 }

.github/workflows/repo-assist.lock.yml

@@ -171,7 +171,7 @@ steps:
           json.dump(result, f, indent=2)
       EOF
 
-source: githubnext/agentics/workflows/repo-assist.md@d1d884596e62351dd652ae78465885dd32f0dd7d
+source: githubnext/agentics/workflows/repo-assist.md@4957663821dbb3260348084fa2f1659701950fef
 ---
 
 # Repo Assist

.github/workflows/repo-assist.md

@@ -1,6 +1,19 @@
 ### 4.11.0
 
 * Performance: `mapiAsync` — replaced `asyncSeq`-builder + `collect` implementation with a direct optimised enumerator (`OptimizedMapiAsyncEnumerator`), eliminating `collect` overhead and bringing per-element cost in line with `mapAsync`. Benchmarks added in `AsyncSeqMapiBenchmarks`.
+* Design parity with FSharp.Control.TaskSeq (#277, batch 2):
+  * Added `AsyncSeq.tryTail` — returns `None` if the sequence is empty; otherwise returns `Some` of the tail. Safe counterpart to `tail`. Mirrors `TaskSeq.tryTail`.
+  * Added `AsyncSeq.where` / `AsyncSeq.whereAsync` — aliases for `filter` / `filterAsync`, mirroring the naming convention in `TaskSeq` and F# 8 collection expressions.
+  * Added `AsyncSeq.lengthBy` / `AsyncSeq.lengthByAsync` — counts elements satisfying a predicate. Mirrors `TaskSeq.lengthBy` / `TaskSeq.lengthByAsync`.
+  * Added `AsyncSeq.compareWith` / `AsyncSeq.compareWithAsync` — lexicographically compares two async sequences using a comparison function. Mirrors `TaskSeq.compareWith` / `TaskSeq.compareWithAsync`.
+  * Added `AsyncSeq.takeWhileInclusiveAsync` — async variant of the existing `takeWhileInclusive`. Mirrors `TaskSeq.takeWhileInclusiveAsync`.
+  * Added `AsyncSeq.skipWhileInclusive` / `AsyncSeq.skipWhileInclusiveAsync` — skips elements while predicate holds and also skips the first non-matching boundary element. Mirrors `TaskSeq.skipWhileInclusive` / `TaskSeq.skipWhileInclusiveAsync`.
+  * Added `AsyncSeq.appendSeq` — appends a synchronous `seq<'T>` after an async sequence. Mirrors `TaskSeq.appendSeq`.
+  * Added `AsyncSeq.prependSeq` — prepends a synchronous `seq<'T>` before an async sequence. Mirrors `TaskSeq.prependSeq`.
+  * Added `AsyncSeq.delay` — defers sequence creation to enumeration time by calling a factory function each time `GetAsyncEnumerator` is called. Mirrors `TaskSeq.delay`.
+  * Added `AsyncSeq.collectAsync` — like `collect` but the mapping function is asynchronous (`'T -> Async<AsyncSeq<'U>>`). Mirrors `TaskSeq.collectAsync`.
+  * Added `AsyncSeq.partition` / `AsyncSeq.partitionAsync` — splits a sequence into two arrays using a (optionally async) predicate; the first array contains matching elements, the second non-matching. Mirrors `TaskSeq.partition` / `TaskSeq.partitionAsync`.
+* Tests: added 14 new unit tests covering previously untested functions — `AsyncSeq.indexed`, `AsyncSeq.iteriAsync`, `AsyncSeq.tryLast`, `AsyncSeq.replicateUntilNoneAsync`, and `AsyncSeq.reduceAsync` (empty-sequence edge case).
 
 ### 4.10.0
 
@@ -11,7 +24,9 @@
 * Performance: `filterAsync` — replaced `asyncSeq`-builder implementation with a direct optimised enumerator, reducing allocation and generator overhead.
 * Performance: `chooseAsync` — fallback (non-`AsyncSeqOp`) path now uses a direct optimised enumerator instead of the `asyncSeq` builder.
 * Performance: `foldAsync` — fallback (non-`AsyncSeqOp`) path now uses a direct loop instead of composing `scanAsync` + `lastOrDefault`, avoiding intermediate sequence allocations.
-* Benchmarks: added `AsyncSeqFilterChooseFoldBenchmarks` and `AsyncSeqPipelineBenchmarks` benchmark classes to measure `filterAsync`, `chooseAsync`, `foldAsync`, `toArrayAsync`, and common multi-step pipelines.
+* Performance: `take` — replaced `asyncSeq`-builder implementation with a direct optimised enumerator (`OptimizedTakeEnumerator`), eliminating generator-machinery overhead for this common slicing operation.
+* Performance: `skip` — replaced `asyncSeq`-builder implementation with a direct optimised enumerator (`OptimizedSkipEnumerator`), eliminating generator-machinery overhead for this common slicing operation.
+* Benchmarks: added `AsyncSeqFilterChooseFoldBenchmarks`, `AsyncSeqPipelineBenchmarks`, and `AsyncSeqSliceBenchmarks` benchmark classes.
 
 ### 4.8.0