schema_version = 1

extends     = "llama.cpp Base Chat"
name        = "llama.cpp Quick Refactor"
description = "Local llama.cpp deterministic inline refactor via the OpenAI-compatible Chat Completions API. Static output rules from :/tasks/quick-refactor.md; QuickRefactorHandler injects the live editor context. Tools off by default for a fast single shot; enabling them needs a tool-capable model + llama-server --jinja."

model        = "qwen2.5-coder-7b-instruct"
enable_tools = false
tags         = ["refactor", "llama.cpp", "local"]

system_prompt = """{{ read_file(":/tasks/quick-refactor.md") }}"""

[body]
max_tokens  = 8192
temperature = 0.2