Pangram Labs Quickstart Guide

Usage

Install pangram

pip install pangram-sdk

Add your API key

Either export it as an environment variable:

export PANGRAM_API_KEY=<your API key>

Or pass it directly to the constructor:

from pangram import Pangram

my_api_key = ''  # Fill this in with your API key.
pangram_client = Pangram(api_key=my_api_key)

Make a request

from pangram import Pangram

pangram_client = Pangram()
result = pangram_client.predict(text)
# Score in range [0, 1] where 0 is human-written and 1 is AI-generated.
score = result['ai_likelihood']

Make a batch request

from pangram import Pangram

text_batch = ["text1", "text2"]

pangram_client = Pangram()
results = pangram_client.batch_predict(text_batch)
for result in results:
    text = result["text"]
    score = result["ai_likelihood"]

Batch Inference

Batch inference has significantly higher throughput, but can incur some startup latency especially if multiple batch requests are sent at once. Use the single inference endpoint if latency is a strong requirement. Use the batch inference endpoint if operating on multiple inputs at once.