📝 document buffered sink methods

docs/api/buffered-sink-flush.md

@@ -0,0 +1,71 @@
+---
+name: buffered-sink-flush
+group: api
+category: sink
+update-time: 20260613
+description: Flush buffered records from a BufferedSink into its wrapped sink.
+key-word:
+    - sink
+    - buffer
+    - flush
+    - public
+---
+
+## Buffered-sink-flush
+
+Flush buffered records from a `BufferedSink[S]` into its wrapped sink. This is the direct sink-level batching control API for simple synchronous buffering.
+
+### Interface
+
+```moonbit
+pub fn[S : Sink] BufferedSink::flush(self : BufferedSink[S]) -> Unit {
+```
+
+#### input
+
+- `self : BufferedSink[S]` - Buffered sink whose pending records should be forwarded to the wrapped sink.
+
+### Explanation
+
+Detailed rules explaining key parameters and behaviors
+
+- If the buffer is empty, the method does nothing.
+- When buffered records exist, they are copied into a temporary `pending` array, the live buffer is cleared, and each record is then written to the wrapped sink in order.
+- This helper drains only the local buffer; any later behavior still depends on the wrapped sink `S`.
+- The method does not return a count or success flag.
+
+### How to Use
+
+Here are some specific examples provided.
+
+#### When Need Explicit Batch Delivery
+
+When a buffered sink should forward pending records before the threshold is reached:
+```moonbit
+sink.flush()
+```
+
+In this example, callers force buffered records to be written to the wrapped sink immediately.
+
+#### When Need A Manual Flush Barrier
+
+When tests or synchronous code want buffering but still need a deliberate release point:
+```moonbit
+let sink = buffered_sink(console_sink(), flush_limit=4)
+sink.flush()
+```
+
+In this example, the buffered sink exposes a direct batching barrier without using queue semantics.
+
+### Error Case
+
+e.g.:
+- If the wrapped sink's write behavior fails or has side effects, this helper does not add an extra reporting layer on top of `S`.
+
+- If callers need a count-returning drain API with drop tracking, `QueuedSink::flush()` or `QueuedSink::drain()` may fit better.
+
+### Notes
+
+1. Use this helper when code owns a `BufferedSink` directly and wants explicit control over when buffered records are forwarded.
+
+2. Automatic flush still happens when buffered record count reaches `flush_limit` during writes.

docs/api/buffered-sink-pending-count.md

@@ -0,0 +1,75 @@
+---
+name: buffered-sink-pending-count
+group: api
+category: sink
+update-time: 20260613
+description: Read the current buffered-record count from a BufferedSink.
+key-word:
+    - sink
+    - buffer
+    - queue
+    - public
+---
+
+## Buffered-sink-pending-count
+
+Read the current buffered-record count from a `BufferedSink[S]`. This is the direct sink-level metric for how many records are still waiting in the in-memory buffer.
+
+### Interface
+
+```moonbit
+pub fn[S] BufferedSink::pending_count(self : BufferedSink[S]) -> Int {
+```
+
+#### input
+
+- `self : BufferedSink[S]` - Buffered sink whose current in-memory backlog should be inspected.
+
+#### output
+
+- `Int` - Current number of buffered records.
+
+### Explanation
+
+Detailed rules explaining key parameters and behaviors
+
+- The return value is `self.buffer.val.length()` at the time of the call.
+- This is a point-in-time metric and may change immediately after it is read.
+- It reflects buffered records that have not yet been flushed to the wrapped sink.
+- This helper does not mutate the sink.
+
+### How to Use
+
+Here are some specific examples provided.
+
+#### When Need Direct Buffer Backlog Visibility
+
+When code is working with a `BufferedSink` value directly and wants to observe pending buffered records:
+```moonbit
+let pending = sink.pending_count()
+```
+
+In this example, callers can inspect the current in-memory backlog without touching the wrapped sink.
+
+#### When Verify Manual Flush Progress
+
+When explicit flush steps should be checked operationally:
+```moonbit
+ignore(sink.flush())
+ignore(sink.pending_count())
+```
+
+In this example, the metric helps verify whether buffered records were forwarded out of the local buffer.
+
+### Error Case
+
+e.g.:
+- This helper does not have a normal failure mode; it only reads current buffer length.
+
+- If callers need overflow-aware backlog semantics instead of simple buffering, `QueuedSink::pending_count()` is the better API.
+
+### Notes
+
+1. Use this helper for direct visibility into simple synchronous buffering.
+
+2. Pair it with `flush()` when inspecting whether batched writes have been forwarded.

docs/api/index.md

@@ -131,6 +131,8 @@ BitLogger API navigation.
 - [split-by-level.md](./split-by-level.md)
 - [buffered-sink.md](./buffered-sink.md)
 - [buffered-sink-type.md](./buffered-sink-type.md)
+- [buffered-sink-pending-count.md](./buffered-sink-pending-count.md)
+- [buffered-sink-flush.md](./buffered-sink-flush.md)
 - [queued-sink.md](./queued-sink.md)
 - [queued-sink-type.md](./queued-sink-type.md)
 - [queue-overflow-policy.md](./queue-overflow-policy.md)