#include "transcript_cleaner.h"

#include <cctype>
#include <string_view>

std::string clean_transcript(std::string transcript) {
  // Qwen3-ASR prefixes raw transcripts with the detected language and a text marker.
  constexpr std::string_view kLanguagePrefix = "language ";
  constexpr std::string_view kMarkers[] = {"<asr_text>", "<astext>"};

  size_t prefix = 0;
  if (transcript.compare(0, 3, "\xef\xbb\xbf") == 0) prefix = 3;
  while (prefix < transcript.size() &&
         std::isspace(static_cast<unsigned char>(transcript[prefix]))) {
    prefix++;
  }
  if (transcript.compare(prefix, kLanguagePrefix.size(), kLanguagePrefix) != 0) {
    return transcript;
  }

  size_t marker_position = std::string::npos;
  std::string_view matched_marker;
  for (std::string_view marker : kMarkers) {
    const size_t position = transcript.find(marker, prefix + kLanguagePrefix.size());
    if (position != std::string::npos &&
        (marker_position == std::string::npos || position < marker_position)) {
      marker_position = position;
      matched_marker = marker;
    }
  }
  if (marker_position == std::string::npos) return transcript;
  return transcript.substr(marker_position + matched_marker.size());
}