classer 0.0.4

High-performance AI classification — <200ms latency, up to 100x cheaper, beats GPT-5 mini accuracy

classer

High-performance AI classification

<200ms latency · up to 100x cheaper · beats GPT-5 mini accuracy · self-calibrating accuracy · no prompt engineering

Installation

pip install classer

Quick Start

import classer

# Single-label classification
result = classer.classify(
    text="I can't log in and need a password reset.",
    labels=["billing", "technical_support", "sales", "spam"]
)
print(result.label)      # "technical_support"
print(result.confidence) # 0.94

# With descriptions for better accuracy
lead = classer.classify(
    text="We need a solution for 500 users, what's your enterprise pricing?",
    labels=["hot", "warm", "cold"],
    descriptions={
        "hot": "Ready to buy, asking for pricing or demo",
        "warm": "Interested but exploring options",
        "cold": "Just browsing, no clear intent"
    }
)
print(lead.label)  # "hot"

# Multi-label tagging
result = classer.tag(
    text="Breaking: Tech stocks surge amid AI boom",
    labels=["politics", "technology", "finance", "sports"],
    threshold=0.3
)
for t in result.labels:
    print(f"{t.label}: {t.confidence}")
# technology: 0.92
# finance: 0.78

Configuration

No API key is needed to get started. To unlock higher rate limits, get an API key from classer.ai/api-keys.

export CLASSER_API_KEY=your-api-key

Or configure programmatically:

from classer import ClasserClient

client = ClasserClient(
    api_key="your-api-key"
)

API Reference

classify(text, labels=None, classifier=None, descriptions=None, speed=None, cache=None)

Classify text into exactly one of the provided labels.

result = classer.classify(
    text="Text to classify",
    labels=["label1", "label2"],  # 1-200 possible labels
    descriptions={"label1": "Description for better accuracy"},
    speed="standard",   # "standard" (default, <1s) or "fast" (<200ms)
    cache=True          # Set to False to bypass cache. Default: True
)

result.label        # Selected label
result.confidence   # 0-1 confidence score
result.tokens       # Total tokens used
result.latency_ms   # Processing time in ms
result.cached       # Whether served from cache

tag(text, labels=None, classifier=None, descriptions=None, threshold=None, speed=None, cache=None)

Multi-label tagging — returns all labels above a confidence threshold.

result = classer.tag(
    text="Text to tag",
    labels=["label1", "label2"],  # 1-200 possible labels
    descriptions={"label1": "Description"},
    threshold=0.3,  # Default: 0.3
    speed="standard",  # "standard" (default, <1s) or "fast" (<200ms)
    cache=True      # Set to False to bypass cache. Default: True
)

for t in result.labels:
    print(f"{t.label}: {t.confidence}")

result.tokens       # Total tokens used
result.latency_ms   # Processing time in ms
result.cached       # Whether served from cache

Error Handling

from classer import ClasserError

try:
    result = classer.classify(text="hello", labels=["a", "b"])
except ClasserError as e:
    print(e.status)  # HTTP status code
    print(e.detail)  # Error detail from API

Documentation

Full API reference and guides at docs.classer.ai.

GitHub

github.com/classer-ai/classer-python

License

MIT