tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-06-26 19:55:04 +00:00
Code Issues Packages Projects Releases Wiki Activity

llama-chat : Do not throw when tool parsing fails (#14012)

Browse Source 
Currently when a model generates output which looks like a tool call,
but is invalid an exception is thrown and not handled, causing the cli
or llama-server to bail. Instead, handle the chat parser exception and
simply return the generated text in such cases.

Signed-off-by: Piotr Stankiewicz <piotr.stankiewicz@docker.com>
This commit is contained in:
Piotr
2025-06-14 18:25:15 +02:00
committed by GitHub
parent 2e42be42bd
commit 3cb203c89f
3 changed files with 10 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
common/chat-parser.cpp
View File

@ -49,6 +49,7 @@ bool common_chat_msg_parser::add_tool_call(const std::string & name, const std::
// LOG_DBG("Tool call arguments:\n\traw: %s\n\tresult: %s\n", arguments.c_str(), tool_call.arguments.c_str());
result_.tool_calls.emplace_back(tool_call);
return true;
}
bool common_chat_msg_parser::add_tool_call(const json & tool_call) {
@ -378,3 +379,7 @@ std::optional<common_chat_msg_parser::consume_json_result> common_chat_msg_parse
/* .is_partial = */ found_healing_marker,
};
}
void common_chat_msg_parser::clear_tools() {
result_.tool_calls.clear();
}

2
common/chat-parser.h
View File

@ -115,4 +115,6 @@ class common_chat_msg_parser {
const std::vector<std::vector<std::string>> & args_paths = {},
const std::vector<std::vector<std::string>> & content_paths = {}
);
void clear_tools();
};

4
common/chat.cpp
View File

@ -1921,7 +1921,9 @@ common_chat_msg common_chat_parse(const std::string & input, bool is_partial, co
} catch (const common_chat_msg_partial_exception & ex) {
LOG_DBG("Partial parse: %s\n", ex.what());
if (!is_partial) {
throw std::runtime_error(ex.what());
builder.clear_tools();
builder.move_to(0);
common_chat_parse_content_only(builder);
}
}
auto msg = builder.result();