LabelerConfig

LabelerConfig(
    name,
    model_id,
    size_b,
    provider,
    ollama_name='',
    mlx_name='',
    base_url='http://localhost:11434/v1',
    description='',
)

Configuration for LLM-based bucket labeling.

Presets available as class attributes

• KEYWORDS: Built-in TF-IDF keywords (no LLM)
• LOW: phi3:mini / Phi-3-mini-4k (3.8B)
• LOW_QWEN: qwen2.5:1.5b (1.5B, fastest)
• MEDIUM: qwen2.5:7b (7B, good balance)
• MEDIUM_LLAMA: llama3.1:8b (8B)
• HIGH: qwen2.5:14b (14B, best local)

Example

config = LabelerConfig.MEDIUM labels = classifier.label_buckets(**config.as_kwargs())

Attributes

Name	Description
base_url	str(object=’’) -> str
description	str(object=’’) -> str
mlx_name	str(object=’’) -> str
ollama_name	str(object=’’) -> str

Methods

Name	Description
as_kwargs	Get kwargs for label_buckets() method.
install_cmd	Get command to install/pull this model.
serve_cmd	Get command to start serving this model.

as_kwargs

LabelerConfig.as_kwargs(use_mlx=False)

Get kwargs for label_buckets() method.

install_cmd

LabelerConfig.install_cmd(use_mlx=False)

Get command to install/pull this model.

serve_cmd

LabelerConfig.serve_cmd(use_mlx=False)

Get command to start serving this model.