llama.cpp
0d491814 - Fix GLM 4.5 parser to correctly handle parallel tool calls

Commit
236 days ago
Fix GLM 4.5 parser to correctly handle parallel tool calls - Add trailing p.space() after tool_close to consume whitespace between calls - Default allow_additional to true per JSON Schema spec (missing additionalProperties should default to true) - Make arg_close newline optional (may not be present before </tool_call>) - Simplify content_prefix pattern for non-reasoning case - Update AGENTS.md with learned insights about parallel tool call parsing All 9 GLM 4.6 needle streaming tests now pass. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Author
Parents
Loading