Skip to content

Label

Label

honestroles.label adds derived labels to job postings using heuristic rules and optional LLM classification.

Modules

  • __init__.py: label_jobs orchestrator and re-exports.
  • heuristic.py: rule-based seniority, role category, and tech stack labeling.
  • llm.py: LLM-based label classification using Ollama.

Public API reference

label_jobs(df: pd.DataFrame, *, use_llm: bool = False, model: str = "llama3", labels: list[str] | None = None, column: str = DESCRIPTION_TEXT, ollama_url: str = "http://localhost:11434", batch_size: int = 8, plugin_labelers: list[str] | None = None, plugin_labeler_kwargs: dict[str, dict[str, object]] | None = None) -> pd.DataFrame

Runs heuristic labeling (label_seniority, label_role_category, label_tech_stack). If use_llm=True, appends label_with_llm using the provided LLM args (model, labels, column, ollama_url, batch_size). If plugin_labelers are provided, registered label plugins are applied after built-in labeling.

label_seniority(df: pd.DataFrame, *, title_column: str = TITLE) -> pd.DataFrame

Adds a seniority column based on job title patterns (intern, junior, mid, senior, staff, principal, lead, director, vp, c_level).

label_role_category(df: pd.DataFrame, *, title_column: str = TITLE, description_column: str = DESCRIPTION_TEXT) -> pd.DataFrame

Adds a role_category column by scanning title and description keywords (engineering, data, design, product, marketing, sales, operations, finance, hr, legal, support).

label_tech_stack(df: pd.DataFrame, *, skills_column: str = SKILLS, description_column: str = DESCRIPTION_TEXT) -> pd.DataFrame

Adds a tech_stack column containing a sorted list of detected terms from the configured _TECH_TERMS list.

label_with_llm(...) -> pd.DataFrame

label_with_llm(
    df: pd.DataFrame,
    *,
    model: str = "llama3",
    labels: list[str] | None = None,
    column: str = DESCRIPTION_TEXT,
    ollama_url: str = "http://localhost:11434",
    batch_size: int = 8,
) -> pd.DataFrame

Uses OllamaClient to classify descriptions and writes llm_labels as a list. If the column is missing or Ollama is unavailable, returns the input DataFrame.

Usage examples

import honestroles as hr

df = hr.read_parquet("jobs_current.parquet")
df = hr.clean_jobs(df)
df = hr.label_jobs(df, use_llm=False)

from honestroles.label import label_with_llm

df = label_with_llm(
    df,
    model="llama3",
    labels=["engineering", "data", "product"],
    ollama_url="http://localhost:11434",
)

Design notes

  • Heuristic labeling uses predefined regex patterns and keyword sets; tune them in heuristic.py if your dataset has domain-specific terms.
  • DEFAULT_LABELS in llm.py is used when no label list is provided.