added gpt-oss:20b model to prompt enhancer node
This commit is contained in:
@@ -15,12 +15,12 @@ class OllamaPromptBooster:
|
||||
"tooltip": "If disabled, the prompt will pass through unchanged."
|
||||
}),
|
||||
"model": (
|
||||
["zephyr:7b-beta", "deepseek-r1:8b", "llama3.2:latest", "mistral:latest"],
|
||||
{
|
||||
"default": "zephyr:7b-beta",
|
||||
"tooltip": "Choose which LLM to use for enhancement. No API key needed."
|
||||
}
|
||||
),
|
||||
["zephyr:7b-beta", "deepseek-r1:8b", "llama3.2:latest", "mistral:latest", "gpt-oss:20b", "gpt-oss:120b"],
|
||||
{
|
||||
"default": "zephyr:7b-beta",
|
||||
"tooltip": "Choose which LLM to use for enhancement. No API key needed."
|
||||
}
|
||||
),
|
||||
"cleanup_output": ("BOOLEAN", {
|
||||
"default": True,
|
||||
"tooltip": "Remove extra formatting, <think> tags, or LLM commentary from output."
|
||||
|
||||
Reference in New Issue
Block a user