Skip to content

[RFC] autoparser unable to accommodate non-ASCII reasoning markers, e.g. ◁think▷ #25369

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
ulidtko wants to merge 1 commit into
base: master
Choose a base branch
from
+30 −1
Open

[RFC] autoparser unable to accommodate non-ASCII reasoning markers, e.g. ◁think▷ #25369

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
27 changes: 27 additions & 0 deletions common/chat-auto-parser-helpers.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -270,6 +270,17 @@ std::vector<segment> segmentize_markers(const std::string & text) {
std::vector<segment> retval;
bool in_marker = false;
char marker_opener = '\0';
// For multi-byte UTF-8 openers (e.g. ◁/▷ used by Kimi-VL-Thinking), marker_opener
// is '\0' and marker_closer_bytes holds the expected closing byte sequence.
std::string marker_closer_bytes;

// UTF-8 byte sequences for ◁ (U+25C1) and ▷ (U+25B7)
static const std::string utf8_triangle_open = "\xe2\x97\x81"; // ◁
static const std::string utf8_triangle_close = "\xe2\x96\xb7"; // ▷

auto starts_with_bytes = [&](size_t pos, const std::string & bytes) -> bool {
return text.compare(pos, bytes.size(), bytes) == 0;
};

auto is_marker_opener = [](char c) -> bool { return c == '<' || c == '['; };
auto is_marker_closer = [](char op, char c) -> bool { return (op == '<' && c == '>') || (op == '[' && c == ']'); };
Expand All @@ -284,12 +295,28 @@ std::vector<segment> segmentize_markers(const std::string & text) {
last_border = cur_pos;
in_marker = true;
marker_opener = text[cur_pos];
} else if (!in_marker && starts_with_bytes(cur_pos, utf8_triangle_open)) {
if (last_border < cur_pos) {
retval.push_back(segment(segment_type::TEXT, text.substr(last_border, cur_pos - last_border)));
}
last_border = cur_pos;
in_marker = true;
marker_opener = '\0';
marker_closer_bytes = utf8_triangle_close;
cur_pos += utf8_triangle_open.size() - 1;
} else if (in_marker && is_marker_closer(marker_opener, text[cur_pos])) {
// no need to check because last_border will always be smaller
retval.push_back(segment(segment_type::MARKER, text.substr(last_border, cur_pos - last_border + 1)));
last_border = cur_pos + 1;
in_marker = false;
marker_opener = '\0';
} else if (in_marker && marker_opener == '\0' && starts_with_bytes(cur_pos, marker_closer_bytes)) {
size_t end = cur_pos + marker_closer_bytes.size();
retval.push_back(segment(segment_type::MARKER, text.substr(last_border, end - last_border)));
last_border = end;
in_marker = false;
marker_closer_bytes.clear();
cur_pos = end - 1;
}
}
if (last_border < text.length()) {
Expand Down
4 changes: 3 additions & 1 deletion common/peg-parser.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1465,7 +1465,9 @@ common_peg_parser common_peg_parser_builder::python_value() {
common_peg_parser common_peg_parser_builder::marker() {
auto sharp_bracket_parser = literal("<") + until(">") + literal(">");
auto square_bracket_parser = literal("[") + until("]") + literal("]");
return choice({ sharp_bracket_parser, square_bracket_parser });
// UTF-8 triangle brackets ◁ (U+25C1) and ▷ (U+25B7), used by e.g. Kimi-VL-Thinking
auto triangle_bracket_parser = literal("\xe2\x97\x81") + until("\xe2\x96\xb7") + literal("\xe2\x96\xb7");
return choice({ sharp_bracket_parser, square_bracket_parser, triangle_bracket_parser });
}

common_peg_parser common_peg_parser_builder::json_member(const std::string & key, const common_peg_parser & p) {
Expand Down