API Reference¶
Coming Soon
Comprehensive API documentation is being developed.
Core Modules¶
triage.preprocess¶
from triage.preprocess import clean_description
clean_text = clean_description("URGENT!!! Login FAILED!!!")
# Returns: "urgent login failed"
triage.model¶
from triage.model import load_vectorizer_and_model, predict_event_type
vectorizer, classifier = load_vectorizer_and_model()
label, probabilities = predict_event_type("Suspicious payroll login email")
triage.cli¶
See CLI Usage for command-line interface documentation.
Function Reference¶
Text Processing¶
clean_description(text: str) -> str¶
Cleans and normalizes incident text.
Parameters:
text(str): Raw incident description
Returns:
- str: Cleaned text (lowercase, normalized)
Model Loading¶
load_vectorizer_and_model() -> Tuple[Vectorizer, Classifier]¶
Loads the saved TF–IDF vectorizer and trained classifier used by the CLI.
Returns:
- Tuple:
(vectorizer, classifier)objects ready for inference
Inference¶
predict_event_type(text: str, top_k: int = 5) -> Tuple[str, Optional[Dict[str, float]]]¶
Predicts the most likely incident label and (optionally) a top-k probability breakdown.
Parameters:
text(str): Incident descriptiontop_k(int): Maximum classes to include in the probability dict
Returns:
- Tuple:
(label, probabilities)wherelabelis a string andprobabilitiesis an optional dict of class → probability
Data Structures¶
Prediction Result¶
Programmatic API returns a tuple. For a structured payload, use the CLI with --json.
CLI Integration¶
For programmatic usage, use JSON mode:
import subprocess
import json
result = subprocess.run(
["nlp-triage", "--json", "incident text"],
capture_output=True,
text=True
)
prediction = json.loads(result.stdout)
For usage examples, see Notebooks Overview.