Last updated for pip package version 0.4.x

​
Installation

Install Commonbase from PyPI using pip:

pip install commonbase

​
Usage

A Project ID and API Key are required for all Commonbase requests. You can find your project ID and generate an API key in the Commonbase Dashboard.

To create a completion, provide your Project ID, API Key, and prompt to Completion.create.

​
Text Completion

To create a basic text completion, use Completion.create with a prompt:

import commonbase

completion = commonbase.Completion.create(
    api_key="API_KEY",
    project_id="PROJECT_ID",
    prompt="Hello!"
)

print(completion.best_choice.text)

completion.best_choice returns the most relevant choice. completion.choices contains the full list of choices returned by the LLM:

for choice in completion.choices:
    print(choice.text)

​
Chat Completion

To create a chat completion, use ChatCompletion.create with a list of messages.

import commonbase

completion = commonbase.ChatCompletion.create(
    api_key="API_KEY",
    project_id="PROJECT_ID",
    messages=[
        { "role": "system", "content": "You are an assistant who helps users with tech problems." },
        { "role": "user", "content": "My internet isn't working." },
        { "role": "assistant", "content": "Have you tried restarting your router?" },
        { "role": "user", "content": "Yes I've tried that." }
    ]
)

print(completion.best_choice.text)

​
Variables

You can also create a completion by providing variables and a template prompt.

import commonbase

completion = commonbase.Completion.create(
    api_key="API_KEY",
    project_id="PROJECT_ID",
    prompt="A new user {{user_name}} just signed up with {{email}}. Say hello!"
    variables={
        "user_name": "Alice",
        "email": "alice@example.com"
    }
)

print(completion.best_choice.text)

​
Streaming

To stream the completion response as it is generated, use ChatCompletion.stream.

Streaming is useful when you want to display a completion “as it is typed” for a chatbot-like experience.

For longer completions, streaming lets you avoid hanging while you wait for a completed response from create.

import commonbase

completionStream = commonbase.ChatCompletion.stream(
    api_key="API_KEY",
    project_id="PROJECT_ID",
    messages=[
        { "role": "user", "content": "Write me a short essay about artificial intelligence." }
    ]
)

for completion in completionStream:
    if not completion.completed:
        print(completion.best_choice.text, end="")
    else:
        # The final response from the stream contains the full completion text.
        # We can ignore it here because we already printed the intermediate results.
        print("\n\ndone")

​
Functions

To use OpenAI’s Function Calling feature, make a call to ChatCompletion.create and provide a list of functions.

import commonbase
import json

# In Production, you'd call a real weather API.
def get_current_weather(location: str, unit: str = "fahrenheit"):
    return {
        "location": location,
        "temperature": 72,
        "unit": unit,
        "forecast": ["sunny", "windy"]
    }

messages: list[commonbase.ChatMessage] = [
    { "role": "user", "content": "What's the weather like in Boston?" }
]

response = commonbase.ChatCompletion.create(
    api_key="API_KEY",
    project_id="PROJECT_ID",
    messages=messages,
    functions=[
        {
            "name": "get_current_weather",
            "description": "Get the current weather in a given location",
            "parameters": {
                "type": "object",
                "properties": {
                    "location": {
                        "type": "string",
                        "description": "The city and state, e.g. San Francisco, CA",
                    },
                    "unit": {"type": "string", "enum": ["celsius", "fahrenheit"]},
                },
                "required": ["location"],
            },
        }
    ]
)

best_choice = response.best_choice

# Check if GPT wanted to call a function
if (best_choice.function_call is not None):

    function_name = best_choice.function_call.name

    # The Commonbase SDK deserializes the arguments for you, though it
    # may not always be valid JSON.
    #
    # If the arguments are empty, check function_call.json["arguments"]
    # to see the raw response from GPT.
    function_arguments = best_choice.function_call.arguments
    location, unit = function_arguments.get("location"), function_arguments.get("unit")

    # Only one function in this example, but you can have multiple.
    available_functions = {
        "get_current_weather": get_current_weather
    }

    function_response = available_functions[function_name](location, unit)

    # Extend the conversation with the assistant's reply.
    messages.append(response.best_choice.to_assistant_chat_message())

    # Extend the conversation with the function response.
    messages.append({
        "role": "function",
        "name": function_name,
        "content": json.dumps(function_response)
    })

    # Get a new response from GPT where it can see the function response.
    second_response = commonbase.ChatCompletion.create(
        api_key="API_KEY",
        project_id="PROJECT_ID",
        messages=messages
    )

    print(second_response.best_choice.text)
JavaScriptC# .NET