Phonic Python SDK

0.31.15 verified Fri May 01 auth: no python

Phonic is an AI evaluation and monitoring platform. The Python SDK (v0.31.15) provides tools to log evaluations, manage datasets, and analyze AI model performance. Release cadence is irregular; breaking changes may occur.

pip install phonic

Common errors

error ImportError: cannot import name 'Phonic' from 'phonic' ↓

cause Old installation or wrong import path; class name changed from 'Client' to 'Phonic'.

fix

Run 'pip install --upgrade phonic' and use 'from phonic import Phonic'.

error AttributeError: module 'phonic' has no attribute 'Client' ↓

cause Using deprecated class name 'Client' which was removed in v0.30.0.

fix

Use 'from phonic import Phonic' and instantiate 'Phonic(api_key=...)'.

error TypeError: log_evaluation() got an unexpected keyword argument 'correct' ↓

cause Parameter 'correct' was renamed to 'expected' in recent versions.

fix

Replace keyword argument 'correct' with 'expected'.

Warnings

breaking The 'Phonic' class replaced 'Client' in v0.30.0; 'Client' is removed. ↓

fix Replace 'from phonic import Client' with 'from phonic import Phonic'.

deprecated Argument 'correct' in log_evaluation is deprecated; use 'expected' instead. ↓

fix Replace 'correct=...' with 'expected=...'.

gotcha API key must be passed as string; environment variable 'PHONIC_API_KEY' is not automatically read by client init. Use os.environ.get() explicitly. ↓

fix client = Phonic(api_key=os.environ.get('PHONIC_API_KEY', ''))

Imports

Phonic
wrong
```
import phonic
```
correct
```
from phonic import Phonic
```
Direct import of module fails; import class explicitly.

EvaluationLog

wrong

from phonic.evaluations import EvaluationLog

correct

from phonic import EvaluationLog

Path changed in recent versions; now top-level.

Quickstart

Initialize client with API key from env var and log an evaluation.

from phonic import Phonic

client = Phonic(api_key=os.environ.get('PHONIC_API_KEY', ''))
# Log an evaluation
client.log_evaluation(
    project_name='my-project',
    input='What is the capital of France?',
    output='Paris',
    expected='Paris',
    scores={'accuracy': 1.0}
)