Accessing Completions and Working with Chat Models
Accessing Completions and Working with Chat Models
In this section, we will explore how to access completions and work with chat models in Keymaker.
Accessing Completions
When you generate completions with Keymaker, you can access them through the completions attribute of a Prompt object. Here’s an example of how to access both named and unnamed completions:
from keymaker import Prompt
from keymaker.models import chatgpt
from keymaker.constraints import RegexConstraint
# Initialize the model
chat_model = chatgpt()
# Create a prompt
prompt = Prompt("The weather is ")
# Generate an unnamed completion
constraint1 = RegexConstraint(pattern=r"sunny|rainy|cloudy")
prompt = await prompt.complete(model=chat_model, constraint=constraint1)
# Generate a named completion
constraint2 = RegexConstraint(pattern=r" and (cold|warm|hot)")
prompt = await prompt.complete(
model=chat_model, constraint=constraint2, name="temperature"
)
# Access the unnamed completion
unnamed_completion = prompt.completions[0]
# Access the named completion
named_completion = prompt.completions.temperature
print(prompt)
print(f"Unnamed completion: {unnamed_completion}")
print(f"Named completion: {named_completion}")
Working with Chat Models
Keymaker provides functionality for using roles with chat models. This can help improve performance by providing the model with a clearer context for the conversation. Here’s an example of using roles with a chat model:
from keymaker import Prompt
from keymaker.models import chatgpt
# Initialize the model
chat_model = chatgpt()
# Create a prompt with roles
prompt = Prompt("""
%system%You are a helpful assistant%/system%
%user%What is the capital of France?%/user%
%assistant%The capital of France is Paris.%/assistant%
""")
# Complete the prompt with the chat model
prompt = await prompt.complete(model=chat_model, name="next_response")
# Print the completed prompt
print(prompt)
By using roles, you can provide a clearer context for the conversation, which may help improve the quality of the generated responses.