% perplexipy(3) Version 1.3.1 | Perplexity AI high level API documentation

PerplexiPy - Perplexity AI high level library

client = PerplexityClient() \
print(client.query('What is the meaning of 42?') \
for result in client.queryStreamable('List of all US presidents'): \
    print(result)

PerplexiPy is a high-level, convenience library for interacting with the Perplexity API from any Python 3.9+ application. The library aims to simplify interactions with Perplexity models by encapsulating all the implementation details of the lower level OpenAI API. All interaction between the code and this library occurs in the form of string and native Python objects, not OpenAPI / Swagger mapped objects.

PerplexiPy implements a combination of the Perplexity AI and the OpenAI outputs. API semantics follow the "literate programming" workflow, and attempt to make the resulting code as simple to follow as possible.

Perplexity AI is an AI-powered search engine that uses natural language processing and machine learning to provide accurate and comprehensive answers to end-user queries. It can be argued that it outperforms OpenAI's offerings in accuracy and responsiveness. PerplexiPy enables the creation of Python programs and tools that leverage the power of Perplexity AI.

These documents encompass all information about PerplexiPy:

• This README file, hosted in the perplexipy GitHub repository project on GitHub and exported to the PyPI project page
• A man page autogenerated from this README file, perplexipy.3
• The complete PerplexiPy API reference on GitHub Pages - https://cime-software.github.io/perplexipy/perplexipy.html
• The PerplexiPy-tutorial.ipynb tutorial notebook

The man page can be generated from the source distribution using this command:

make manpage

The output will reside in: ./manpages/perplexipy.3. The man page isn't included in the Python package wheel because there's no accepted standard installation that's cross-compatible among all supported operating systems and distributions.

pip install perplexipy

Package information: https://pypi.org/project/perplexipy

PerplexiPy requires an API key issued by Perplexity. You may set PERPLEXITY_API_KEY as an environment variable or use a .env file.

Access to the Perplexity API requires a paid subscription and an API key.

Use of .env is recommended for storing the private key. There is an automatic constant PERPLEXITY_API_KEY that gets initialized to the value of a .env key of the same name via the dotenv API. Otherwise, PERPLEXITY_API_KEY may be handled like any other secret by the implementing team. This module provides no other tools or services for handling the API key.

This key is used only during PerplexityClient instantiation:

import os

key = os.environ['PERPLEXITY_API_KEY']
client = PerplexityClient(key = key)
print(client.query('Brief answer:  greet the world in Swedish.'))

See the PerplexiPy-tutorial.ipynb notebook for a richer example of the API's capabilities.

In general:

client = PerplexityClient()
result = client.query('Show me how to declare a list in Python')
# result is a string

results = client.queryBatch('Show me different ways of declaring a Python list')
for result in results:
    print(result)
    # results is a tuple of one or more results

results = client.queryStreamable('Tell me why lists are important in programming')
for result in results:
    print(result)
    # results is a long stream of data, served over time.

models = client.models # lists all models supported by Perplexity AI

print('Model names:')
for model in models.keys():
    print(' - %s' % model)

try:
    client.model = 'bogus-LLM-7b'
except PerplexityClientError as e:
    print(e)

client.model = models.keys()[0] # OK

Not supported at this time.

PerplexiPy ships with the Codex Playground, an interactive REPL console. To run it:

# In a virtualenv:
pip install -U perplexipy
export PERPLEXITY_API_KEY="your-key-goes-here"
codex repl

Full description of Codex Playground and other use cases like streaming API and CLI argument passing in the Codex README.

The PerplexiPy package, documentation and examples are licensed under the BSD-3 open source license.

• codex - a PerplexiPy command line code explainer
• API documentation: https://cime-software.github.io/perplexipy
• Perplexity AI
• m0toko: a Sopel bot plug-in that uses PerplexiPy for providing AI-enhanced chatbot functionality: https://github.com/pr3d4t0r/m0toko
• PyPI: openai

The code should work with Python 3.7 or later, but it was only tested with Python 3.9.16 and later. Download the package and install it from source if support for an earlier Python version is required.

Feature requests and bug reports:

https://github.com/CIME-Software/perplexipy/issues