OutboundSender.sendMarkdown/sendText drop replyTo on chunked pages (pages 2+ land in the main chat instead of the thread)

[tkwkeven]

Summary

OutboundSender.sendMarkdown and sendText chunk long content and send each chunk as a separate message. Only the first chunk carries opts.replyTo; chunks 2+ are sent as plain new messages, so a long reply that should live inside a thread (e.g. reply_in_thread: true) leaks into the main chat.

Location

channel/outbound/sender.ts:

// sendMarkdown — chunk loop
replyTo: i === 0 ? opts.replyTo : undefined,

// sendText — chunk loop
replyTo: i === 0 ? opts.replyTo : undefined,

Still present in @larksuiteoapi/node-sdk@1.73.0 (latest; lib/index.js lines ~104701 / ~104721).

Expected behavior

Every chunked page should carry replyTo (and replyInThread), so all pages stay in the same thread. Mentions are already correctly limited to the first page via convertMarkdown(chunks[i], i === 0 ? opts.mentions : undefined) — that behavior should be preserved.

Suggested fix

-            replyTo: i === 0 ? opts.replyTo : undefined,
+            replyTo: opts.replyTo,

(two occurrences: sendMarkdown and sendText)

Verification

We applied this exact two-line patch in production (patching the bundled copy in a downstream consumer) and confirmed: all chunked pages now arrive inside the thread, first page keeps mentions, no other regressions.

Repro sketch

1. createLarkChannel({ outbound: { textChunkLimit: <small> } })
2. Send a markdown message longer than the chunk limit with { replyTo: <msgId>, replyInThread: true }
3. Observe: chunk 1 replies to <msgId>; chunks 2+ are created as top-level messages in the chat.

[mazhe-nerd]

Channel-related capabilities can use channel-sdk：https://github.com/larksuite/channel-sdk-node