Python SDK
The GPTfake Python SDK is the official client for the GPTfake API. It handles authentication, retries, and pagination so you can pull live censorship and bias data — refusal rates, bias scores, history, and datasets — in a few lines.
Install
pip install gptfake-sdkRequires Python 3.8+.
Authenticate
Pass an API key to the client. Load it from an environment variable rather than hardcoding it:
import os
from gptfake import GPTfakeClient
client = GPTfakeClient(os.environ["GPTFAKE_API_KEY"])Never commit API keys to source control. Use environment variables or a secrets manager. See API keys for storage and rotation guidance.
Fetch current metrics
metrics = client.get_metrics('chatgpt')
print(f"Overall refusal rate: {metrics.overall_refusal_rate}%")
print(f"Political refusal rate: {metrics.political_refusal_rate}%")
print(f"Bias score: {metrics.bias_score}")Valid model ids: chatgpt, claude, gemini, mistral, qwen — see Models.
Historical trends
history = client.get_history('chatgpt', period='30d', category='political')
for point in history.series:
print(point.date, point.refusal_rate)period accepts 7d, 30d, 90d, 1y; category accepts political, safety, social, all. See the Metrics endpoint.
Compare models
comparison = client.compare(['chatgpt', 'claude', 'gemini'])
for model in comparison.models:
print(model.id, model.overall_refusal_rate)Export research data
datasets = client.list_datasets()
client.download_dataset('ds_q4_2024', path='gptfake_q4_2024.csv', format='csv')See the Data export reference.
Error handling
from gptfake import GPTfakeError, RateLimitError
try:
metrics = client.get_metrics('chatgpt')
except RateLimitError as e:
print(f"Rate limited, retry after {e.retry_after}s")
except GPTfakeError as e:
print(f"API error: {e.code} — {e.message}")Next steps
- JavaScript SDK
- Monitoring API reference
- Back to the API overview
GPTfake