v0.2.3

• Python package for chaincrafter updated to include the install_requires so that PyYAML is installed correctly when chaincrafter is a dependency

v0.2.1

What's Changed

• Python: rename math.py example file to avoid conflict with math module by @rudolfolah in #10
• Load chains and prompts from YAML using Catalogs by @rudolfolah in #11

Full Changelog: v0.2.0...v0.2.1

Catalogs

A catalog is a collection of chains and prompts stored in a YAML file.

• Catalogs API
• YAML Format for Catalogs of Prompts
• Hello World, loading prompts and chains from a YAML file
• Example of loading prompts and constructing new prompts with prompt modifiers

v0.2.0

What's Changed

Documentation

• Add API and Examples to the docs by @rudolfolah in #6

Python

• fixes #4 passing starting input vars to all prompts in a chain by @rudolfolah in #5
• Testing MockChat with prompts by @rudolfolah in #7

Full Changelog: v0.1.0...v0.2.0

initial release: Python

The initial version of Chaincrafter has been released and published for Python. It currently supports OpenAI, with plans to support local LLMs such as gpt4all and llama.cpp.

Using OpenAI was simplest to test out the asynchronous support with asyncio and the experiments.

Async support

Async lets you make multiple requests to the LLM at the same time:

Example Code

import asyncio

from chaincrafter import Chain, Prompt
from chaincrafter.models import OpenAiChat

chat_model = OpenAiChat(
    temperature=0.9,
    model_name="gpt-3.5-turbo",
    presence_penalty=0.1,
    frequency_penalty=0.2,
)


def make_chain(country):
    system_prompt = Prompt("You are a helpful assistant who responds to questions about the world")
    followup_prompt = Prompt("{city} sounds like a nice place to visit. What is the population of {city}?")
    hello_prompt = Prompt(f"Hello, what is the capital of {country}? Answer only with the city name.")
    return Chain(
        system_prompt,
        (hello_prompt, "city"),
        (followup_prompt, "followup_response"),
    )


async def main():
    chain_france = make_chain("France")
    chain_china = make_chain("China")
    results = await asyncio.gather(
        chain_france.async_run(chat_model),
        chain_china.async_run(chat_model),
    )
    for messages in results:
        for message in messages:
            print(f"{message['role']}: {message['content']}")

asyncio.run(main())

Experiments support

Experiments allow you to test combinations of model parameters with the same prompt. You can use this to compare models, model parameters and to compare them over time.

Example Code

from chaincrafter import Chain, Prompt
from chaincrafter.experiments import OpenAiChatExperiment

system_prompt = Prompt("You are a helpful assistant who responds to questions about the world")
hello_prompt = Prompt("Hello, what is the capital of France? Answer only with the city name.")
followup_prompt = Prompt("{city} sounds like a nice place to visit. What is the population of {city}?")
chain = Chain(
    system_prompt,
    (hello_prompt, "city"),
    (followup_prompt, "followup_response"),
)
experiment = OpenAiChatExperiment(
    chain,
    model_name=["gpt-4", "gpt-3.5-turbo"],
    temperature=[0.7, 1.5],
    presence_penalty=[0.1],
    frequency_penalty=[0.2],
)
experiment.run()
print(experiment.results)
# CSV Output
print(experiment.to_csv())
# JSON Output
print(experiment.to_json())
# Pandas DataFrame Output
print(experiment.to_pandas_df())
# Pandas DataFrame Visualize
print(experiment.visualize())

Links

• Documentation
• Python package at PyPi