schema_version = 1

extends     = "Ollama Base Chat"
name        = "Ollama Compression — 8 GB"
description = "Conversation summarization tuned for ~8 GB RAM — qwen3.5:4b at num_ctx 8192 (~5.5 GB resident). Smallest qwen3.5 (same family as the bigger tiers), decent context for the footprint. KV cache is reserved at load per num_ctx. No tools."

model        = "qwen3.5:4b"
enable_tools = false
tags         = ["compression", "ollama", "local", "8gb"]

system_prompt = """{{ read_file(":/tasks/chat-compressor.md") }}"""

[body.options]
num_predict = 2048
temperature = 0.3
num_ctx     = 8192
keep_alive  = "5m"