Avoid aborting ASR word merge on out-of-span timestamps

src/framework/audio/chunking.cpp

@@ -1,12 +1,14 @@
 #include "engine/framework/audio/chunking.h"
 
+#include "engine/framework/debug/trace.h"
 #include "engine/framework/runtime/options.h"
 
 #include <algorithm>
 #include <cmath>
 #include <cstddef>
 #include <limits>
 #include <optional>
+#include <sstream>
 #include <stdexcept>
 #include <string>
 #include <utility>
@@ -599,7 +601,20 @@ void append_chunk_word_timestamps(
         const int64_t local_start = std::max<int64_t>(word.span.start_sample, 0);
         const int64_t local_end = std::min<int64_t>(word.span.end_sample, source_samples);
         if (local_start >= local_end) {
-            throw std::runtime_error("Audio chunker word merge received a timestamp outside the chunk span");
+            std::ostringstream warning;
+            warning << "dropping word timestamp outside chunk span"
+                    << " word=\"" << word.word << "\""
+                    << " local_start=" << word.span.start_sample
+                    << " local_end=" << word.span.end_sample
+                    << " source_samples=" << source_samples
+                    << " source_start=" << source_span.start_sample
+                    << " source_end=" << source_span.end_sample
+                    << " keep_start=" << keep_span.start_sample
+                    << " keep_end=" << keep_span.end_sample
+                    << " source_sample_rate=" << source_sample_rate
+                    << " timestamp_sample_rate=" << timestamp_sample_rate;
+            debug::log_message(debug::LogLevel::Warning, "audio.chunking", warning.str());
+            continue;
         }
         const int64_t global_start = timestamp_source_span.start_sample + local_start;
         if (global_start < timestamp_keep_span.start_sample || global_start >= timestamp_keep_span.end_sample) {

tests/unittests/test_audio_chunking.cpp

@@ -836,16 +836,6 @@ void test_chunk_word_timestamp_merge_rejects_invalid_spans() {
         },
         "inverted chunk span");
 
-    require_throws(
-        []() {
-            std::vector<engine::runtime::WordTimestamp> merged;
-            engine::audio::append_chunk_word_timestamps(
-                merged,
-                {word("outside", 120, 140)},
-                engine::runtime::TimeSpan{1000, 1100});
-        },
-        "word outside chunk");
-
     require_throws(
         []() {
             std::vector<engine::runtime::WordTimestamp> merged;
@@ -871,6 +861,24 @@ void test_chunk_word_timestamp_merge_rejects_invalid_spans() {
         "invalid source sample rate");
 }
 
+void test_chunk_word_timestamp_merge_drops_outside_words() {
+    std::vector<engine::runtime::WordTimestamp> merged;
+    engine::audio::append_chunk_word_timestamps(
+        merged,
+        {
+            word("kept", 20, 40),
+            word("outside", 120, 140),
+            word("also_kept", 50, 80),
+        },
+        engine::runtime::TimeSpan{1000, 1100});
+
+    engine::test::require_eq(merged.size(), static_cast<size_t>(2), "outside word dropped");
+    engine::test::require_eq(merged[0].word, std::string("kept"), "first valid word kept");
+    require_span(merged[0].span, 1020, 1040, "first valid word span");
+    engine::test::require_eq(merged[1].word, std::string("also_kept"), "second valid word kept");
+    require_span(merged[1].span, 1050, 1080, "second valid word span");
+}
+
 }  // namespace
 
 int main() {
@@ -903,6 +911,7 @@ int main() {
         test_chunk_word_timestamp_merge_rescales_chunk_domain();
         test_chunk_speech_metadata_merge_rescales_chunk_domain();
         test_chunk_word_timestamp_merge_rejects_invalid_spans();
+        test_chunk_word_timestamp_merge_drops_outside_words();
         std::cout << "audio_chunking_test passed\n";
     } catch (const std::exception & ex) {
         std::cerr << "audio_chunking_test failed: " << ex.what() << "\n";