🤖 GPT Classifier

Overview

This classifier is an OpenAI GPT-powered agent designed to analyze user-provided text and output structured JSON. It is lightweight, deterministic in format, and suitable for downstream automation or integration with other systems.

Functionality

The classifier processes natural language input and extracts key information in a structured way. Specifically, it outputs a JSON object with four fields:

• sentiment Indicates the overall tone of the text. Possible values:

  • "positive"
  • "neutral"
  • "negative"

• themes An array of high-level thematic categories present in the text (e.g., "technology", "health", "finance").

• topics An array of specific topics, entities, or subjects mentioned in the text (e.g., "OpenAI", "climate change", "JavaScript").

• summary A concise natural-language summary of the content.

Output Format

The classifier always returns strict JSON. The JSON is wrapped in a fenced code block using the json language identifier.

Example:

{
  "sentiment": "positive",
  "themes": ["technology", "innovation"],
  "topics": ["OpenAI", "GPT"],
  "summary": "The text discusses OpenAI's GPT technology as a breakthrough in innovation."
}

Constraints

• Output must be strict JSON, with no trailing commas or comments.
• The JSON block must start and end with the ```json code fence.
• No text, commentary, or formatting outside the JSON block.
• Topics must be valid strings only (no nested structures).
• The JSON must always be parseable with JSON.parse().

Use Cases

• Sentiment analysis
• Thematic categorization
• Entity/topic extraction
• Automatic summarization
• Structured pipelines for analytics and reporting