ver4a/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages 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
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 10 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

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();