Overview

The Classifier Service is a specialized service designed to classify text into predefined categories using LLMs. It supports both single-class and multi-class classification, with configurable allowed classes and classification types, making it highly flexible for various classification tasks.

• Single and Multi-Class Classification: Support for both single and multiple category assignments
• Configurable Categories: Define allowed classes through enums
• Context-Aware Classification: Domain-specific classification with type specification
• Configurable Output: Control over classification parameters and model behavior
• Confidence Scoring: Detailed confidence metrics for classifications
• Explanatory Reasoning: Comprehensive explanation for classification decisions

Traditional text classification often involves rule-based systems, statistical models, or manual category mapping. The Classifier Service improves this by using LLMs for intelligent classification, supporting multiple classification modes, providing detailed reasoning, and maintaining high accuracy through controlled parameters.

Endpoints

Process Endpoint

The process endpoint performs text classification with AI-powered category assignment.

API Endpoint

Endpoint Parameters

Request

content (string, required)

The text to classify.

context (object, required)

Classification type and allowed classes:

• classificationType (string, optional): Domain specification for classification
• allowedClasses (string[]): Valid category list to classify into
• multiClass (boolean, optional): Whether to allow multiple class assignments (default: false)

config (object, optional)

Model and classification configuration:

• modelId (string): Model to use for classification (default: "mistral-large-2411")
• temperature (number): Controls classification precision (default: 0.1, range: 0.0-1.0)

Response

status (string)

Status of the processing ("success" or "error").

result (object)

For successful responses:

• class (string): Assigned class (single-class mode)
• classes (string[]): Assigned classes (multi-class mode)
• confidence (number): Confidence score for the classification
• explanation (string): Detailed reasoning for the classification decision

metadata (object)

Processing metadata:

• model (string): Model used
• created_at (string): Creation timestamp
• processed_at (string): Processing timestamp
• instance_id (string): Instance identifier
• tracking_id (string): Unique tracking ID
• classification_type (string): Type of classification performed
• mode (string): Classification mode used ("single_class" or "multi_class")

Examples

Request

Full Request (Single-Class Classification)

Multi-Class Classification

Domain-Specific Classification

Sentiment Classification

Content Category Classification

Language Detection

Response

Single-Class Response

Multi-Class Response

Error Response

Classification Modes

Single-Class Mode

• One class per input
• Higher confidence threshold
• Strict class selection

Multi-Class Mode

• Multiple classes per input
• Confidence per class
• Class relationship analysis