Classify into a category tree

Open in playground →

Pick a department first, then a category inside it, in one call. Each step chooses from a short list, so the tree can hold far more categories than one list.

Enum
when

Challenges

department chooses from 3 departments. Each department has its own field with its categories, and "when": {"department": ...} runs only the one that matches, so the other two are skipped.
A French press goes to home_kitchen, then coffee_tea. A backpacking tent, "Two-person backpacking tent, 4 lb, waterproof rainfly.", came back sports_outdoors and camping, and the other two departments were skipped.

import os
from typellm import TypeLLMClient

client = TypeLLMClient(api_key=os.environ["TYPELLM_API_KEY"])
response = client.generate(
    context="Product: Stainless steel French press, 34 oz, double-wall insulated, makes 8 cups.",
    questions={
      "department": {
        "type": "string",
        "enum": [
          "electronics",
          "home_kitchen",
          "sports_outdoors"
        ],
        "instructions": "Which department does this product belong to?"
      },
      "electronics": {
        "type": "string",
        "enum": [
          "phones",
          "laptops",
          "audio",
          "cameras"
        ],
        "instructions": "Which kind of electronics is it?",
        "when": {
          "department": "electronics"
        }
      },
      "home_kitchen": {
        "type": "string",
        "enum": [
          "cookware",
          "coffee_tea",
          "bedding",
          "lighting"
        ],
        "instructions": "Which kind of home and kitchen product is it?",
        "when": {
          "department": "home_kitchen"
        }
      },
      "sports_outdoors": {
        "type": "string",
        "enum": [
          "camping",
          "cycling",
          "fitness",
          "hiking"
        ],
        "instructions": "Which kind of sports and outdoors product is it?",
        "when": {
          "department": "sports_outdoors"
        }
      }
    },
)
print(response)

Time 0.21 s · Cost $0.000011

Generation(
    result={
        'department': 'home_kitchen',
        'home_kitchen': 'coffee_tea',
    },
    thinking={},
    usage=Usage(input_tokens=218, thinking_tokens=0),
    skipped=[
        'electronics',
        'sports_outdoors',
    ],
)