Welcome to prompttools created by Hegel AI! This repo offers a set of free, open-source tools for testing and experimenting with prompts. The core idea is to enable developers to evaluate prompts using familiar interfaces like code and notebooks.
To install prompttools, you can use pip:
pip install prompttools
You can run a simple example of a prompttools with the following
DEBUG=1 python examples/prompttests/example.py
To run the example outside of DEBUG mode, you'll need to bring your own OpenAI API key.
This is because prompttools makes a call to OpenAI from your machine. For example:
OPENAI_API_KEY=sk-... python examples/prompttests/example.py
You can see the full example here.
There are primarily two ways you can use prompttools in your LLM workflow:
- Run experiments in notebooks.
- Write unit tests and integrate them into your CI/CD workflow via Github Actions.
There are a few different ways to run an experiment in a notebook.
The simplest way is to define an experimentation harness and an evaluation function:
def eval_fn(prompt: str, results: Dict, metadata: Dict) -> float:
# Your logic here, or use a built-in one such as `prompttools.utils.similarity`.
pass
prompt_templates = [
"Answer the following question: {{input}}",
"Respond the following query: {{input}}"
]
user_inputs = [
{"input": "Who was the first president?"},
{"input": "Who was the first president of India?"}
]
harness = PromptTemplateExperimentationHarness("text-davinci-003",
prompt_templates,
user_inputs)
harness.run()
harness.evaluate("metric_name", eval_fn)
harness.visualize() # The results will be displayed as a table in your notebook
If you are interested to compare different models, the ModelComparison example may be of interest.
For an example of built-in evaluation function, please see this example of semantic similarity comparison for details.
You can also manually enter feedback to evaluate prompts, see HumanFeedback.ipynb.
Note: Above we used an
ExperimentationHarness. Under the hood, that harness uses anExperimentto construct and make API calls to LLMs. The harness is responsible for managing higher level abstractions, like prompt templates or system prompts. To see how experiments work at a low level, see this example.
Unit tests in prompttools are called prompttests. They use the @prompttest annotation to transform an evaluation function into an efficient unit test. The prompttest framework executes and evaluates experiments so you can test prompts over time. You can see an example test here and an example of that test being used as a Github Action here.
To persist the results of your tests and experiments, one option is to enable HegelScribe (also developed by us at Hegel AI). It logs all the inferences from your LLM, along with metadata and custom metrics, for you to view on your private dashboard. We have a few early adopters right now, and
we can further discuss your use cases, pain points, and how it may be useful for you.
To install prompttools using pip:
pip install prompttoolsTo install from source, first clone this GitHub repo to your local machine, then, from the repo, run:
pip install .You can then proceed to run our examples.
- Will this library forward my LLM calls to a server before sending it to OpenAI/Anthropic/etc?
- No, the source code will be executed on your machine. Any call to LLM APIs will be directly executed from your machine without any forwarding.
We welcome PRs and suggestions! Don't hesitate to open a PR/issue or to reach out to us via email.
We will be delighted to work with early adopters to shape our designs. Please reach out to us via email if you're interested in using this tooling for your project or have any feedback.
We will be gradually releasing more components to the open-source community. The current license can be found in the LICENSE file. If there is any concern, please contact us and we will be happy to work with you.