Skip to content

Solution for long term memory for agent coding CLIs and to facilitate handoff between different agent vendors

Rust

fix(importer): keep subtitle cues that follow two blank lines

Max Freedom Pollarde0a4570b7f25
diff --git a/CHANGELOG.md b/CHANGELOG.md
index e2ce2b7a..2f35e8e7 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -20,6 +20,14 @@ version in [`docs/changelog/`](docs/changelog/):
does not run on this machine" for a release that runs from where it is
installed. A binary that really does not run is still refused, naming the
directory it was tried from, and its staged copy is removed (#1623).
+- The document importer no longer drops a VTT or SRT cue that follows two
+ blank lines (or any even number of them). The subtitle reader split cues on
+ a single blank line, so the extra one was left at the top of the next block,
+ and a block that started with an empty line was skipped whole. A WebVTT file
+ whose header was followed by two blank lines lost its first cue, or was
+ rejected as containing no extractable text when that was its only cue.
+ WebVTT allows any number of blank lines between blocks; the reader now skips
+ them before reading a block.
## [3.2.2] - 2026-10-22
diff --git a/companions/ai-memory-importer/README.md b/companions/ai-memory-importer/README.md
index 7068737e..ea8fe0d4 100644
--- a/companions/ai-memory-importer/README.md
+++ b/companions/ai-memory-importer/README.md
@@ -145,8 +145,9 @@ by formatting may still be extracted; deleted revisions and field instructions
are excluded. Review `--show-body` before every live import of a document whose
visibility rules matter.
-VTT/SRT files accept LF, CRLF, or CR line endings. WebVTT comments are omitted;
-identifiers such as `NOTE123` keep their cue text.
+VTT/SRT files accept LF, CRLF, or CR line endings and any number of blank lines
+between cues. WebVTT comments are omitted; identifiers such as `NOTE123` keep
+their cue text.
## Safety contract
diff --git a/companions/ai-memory-importer/src/document.rs b/companions/ai-memory-importer/src/document.rs
index 063f1908..3622194b 100644
--- a/companions/ai-memory-importer/src/document.rs
+++ b/companions/ai-memory-importer/src/document.rs
@@ -1029,8 +1029,15 @@ fn extract_transcript(bytes: &[u8], webvtt: bool) -> Result<String> {
.replace("\r\n", "\n")
.replace('\r', "\n");
for (block_index, block) in normalized.split("\n\n").enumerate() {
- let lines = block.lines().map(str::trim).collect::<Vec<_>>();
- let Some(first) = lines.first().copied().filter(|line| !line.is_empty()) else {
+ // Blocks can be separated by more than one blank line (WebVTT allows
+ // any number: https://www.w3.org/TR/webvtt1/#file-structure), so
+ // splitting on a single one can leave a blank line atop the next block.
+ let lines = block
+ .lines()
+ .map(str::trim)
+ .skip_while(|line| line.is_empty())
+ .collect::<Vec<_>>();
+ let Some(first) = lines.first().copied() else {
continue;
};
if webvtt && block_index == 0 && first.starts_with("WEBVTT") {
@@ -1752,6 +1759,22 @@ mod tests {
}
}
+ #[test]
+ fn transcripts_keep_cues_after_more_than_one_blank_line() {
+ let vtt = "WEBVTT\n\n\n00:00:01.000 --> 00:00:02.000\nFirst fact\n\n\n00:00:03.000 --> 00:00:04.000\nSecond fact\n";
+ let srt = "1\n00:00:01,000 --> 00:00:02,000\nFirst fact\n\n\n2\n00:00:03,000 --> 00:00:04,000\nSecond fact\n";
+ for newline in ["\n", "\r\n", "\r"] {
+ for (input, webvtt) in [(vtt, true), (srt, false)] {
+ let input = input.replace('\n', newline);
+ assert_eq!(
+ extract_transcript(input.as_bytes(), webvtt).unwrap().trim(),
+ "First fact\n\nSecond fact",
+ "line terminator {newline:?}, WebVTT={webvtt}"
+ );
+ }
+ }
+ }
+
#[test]
fn subtitle_cues_are_sanitized_before_the_document_plan() {
let temp = TempDir::new().unwrap();