Files
QodeAssist/sources/agents/ollama_compression_32gb.toml
2026-06-29 10:05:28 +02:00

18 lines
562 B
TOML

schema_version = 1
extends = "Ollama Base Chat"
name = "Ollama Compression — 32 GB"
description = "Conversation summarization tuned for ~32 GB RAM — qwen3.5:9b at num_ctx 24576 (~20 GB resident). Long context for big chats. KV cache is reserved at load per num_ctx. No tools."
model = "qwen3.5:9b"
enable_tools = false
tags = ["compression", "ollama", "local", "32gb"]
system_prompt = """{{ read_file(":/tasks/chat-compressor.md") }}"""
[body.options]
num_predict = 2048
temperature = 0.3
num_ctx = 24576
keep_alive = "5m"