not-again-ai 0.13.0

Designed to once and for all collect all the little things that come up over and over again in AI projects and put them in one place.

not-again-ai

not-again-ai is a collection of various building blocks that come up over and over again when developing AI products. The key goals of this package are to have simple, yet flexible interfaces and to minimize dependencies. It is encouraged to also a) use this as a template for your own Python package. b) instead of installing the package, copy and paste functions into your own projects. We make this easier by limiting the number of dependencies and use an MIT license.

Documentation available within individual notebooks, docstrings within the source, or auto-generated at DaveCoDev.github.io/not-again-ai/.

Installation

Requires: Python 3.11, or 3.12

Install the entire package from PyPI with:

$ pip install not_again_ai[llm,local_llm,statistics,viz]

Note that local LLM requires separate installations and will not work out of the box due to how hardware dependent it is. Be sure to check the notebooks for more details.

The package is split into subpackages, so you can install only the parts you need.

• Base only: pip install not_again_ai
• LLM: pip install not_again_ai[llm]
  1. OpenAI API
     1. Go to https://platform.openai.com/settings/profile?tab=api-keys to get your API key.
     2. (Optional) Set the OPENAI_API_KEY and the OPENAI_ORG_ID environment variables.
  2. Azure OpenAI (AOAI)
     1. Using AOAI requires using Entra ID authentication. See https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/managed-identity for how to set this up for your AOAI deployment.
        1. Requires the correct role assigned to your user account and being signed into the Azure CLI.
     2. (Optional) Set the AZURE_OPENAI_ENDPOINT environment variable.
  3. GitHub Models
     1. Get a Personal Access Token from https://github.com/settings/tokens and set the GITHUB_TOKEN environment variable. The token does not need any permissions.
     2. Check the Github Marketplace to see which models are available.
• Local LLM: pip install not_again_ai[llm,local_llm]
  1. Some HuggingFace transformers tokenizers are gated behind access requests. If you wish to use these, you will need to request access from HuggingFace on the model card.
     1. Then set the HF_TOKEN environment variable to your HuggingFace API token which can be found here: https://huggingface.co/settings/tokens
  2. If you wish to use Ollama:
     1. Follow the instructions at https://github.com/ollama/ollama to install Ollama for your system.
     2. (Optional) Add Ollama as a startup service (recommended)
     3. (Optional) To make the Ollama service accessible on your local network from a Linux server, add the following to the /etc/systemd/system/ollama.service file which will make Ollama available at http://<local_address>:11434:

        [Service]
        ...
        Environment="OLLAMA_HOST=0.0.0.0"
        

     4. It is recommended to always have the latest version of Ollama. To update Ollama check the docs. The command for Linux is: curl -fsSL https://ollama.com/install.sh | sh
  3. HuggingFace transformers and other requirements are hardware dependent so for providers other than Ollama, this only installs some generic dependencies. Check the notebooks for more details on what is available and how to install it.
• Statistics: pip install not_again_ai[statistics]
• Visualization: pip install not_again_ai[viz]

Development Information

The following information is relevant if you would like to contribute or use this package as a template for yourself.

This package uses Poetry to manage dependencies and isolated Python virtual environments. To proceed, be sure to first install pipx and then install Poetry.

Install Poetry Plugin: Export

$ pipx inject poetry poetry-plugin-export

(Optional) configure Poetry to use an in-project virtual environment.

$ poetry config virtualenvs.in-project true

Dependencies

Dependencies are defined in pyproject.toml and specific versions are locked into poetry.lock. This allows for exact reproducible environments across all machines that use the project, both during development and in production.

To upgrade all dependencies to the versions defined in pyproject.toml:

$ poetry update

To install all dependencies (with all extra dependencies) into an isolated virtual environment:

Append --sync to uninstall dependencies that are no longer in use from the virtual environment.

$ poetry install --all-extras

To activate the virtual environment that is automatically created by Poetry:

$ poetry shell

To deactivate the environment:

(.venv) $ exit

Packaging

This project is designed as a Python package, meaning that it can be bundled up and redistributed as a single compressed file.

Packaging is configured by:

• pyproject.toml

To package the project as both a source distribution and a wheel:

$ poetry build

This will generate dist/not-again-ai-<version>.tar.gz and dist/not_again_ai-<version>-py3-none-any.whl.

Read more about the advantages of wheels to understand why generating wheel distributions are important.

Publish Distributions to PyPI

Source and wheel redistributable packages can be published to PyPI or installed directly from the filesystem using pip.

$ poetry publish

Enforcing Code Quality

Automated code quality checks are performed using Nox and nox-poetry. Nox will automatically create virtual environments and run commands based on noxfile.py for unit testing, PEP 8 style guide checking, type checking and documentation generation.

Note: nox is installed into the virtual environment automatically by the poetry install command above. Run poetry shell to activate the virtual environment.

To run all default sessions:

(.venv) $ nox

Unit Testing

Unit testing is performed with pytest. pytest has become the de facto Python unit testing framework. Some key advantages over the built-in unittest module are:

1. Significantly less boilerplate needed for tests.
2. PEP 8 compliant names (e.g. pytest.raises() instead of self.assertRaises()).
3. Vibrant ecosystem of plugins.

pytest will automatically discover and run tests by recursively searching for folders and .py files prefixed with test for any functions prefixed by test.

The tests folder is created as a Python package (i.e. there is an __init__.py file within it) because this helps pytest uniquely namespace the test files. Without this, two test files cannot be named the same, even if they are in different subdirectories.

Code coverage is provided by the pytest-cov plugin.

When running a unit test Nox session (e.g. nox -s test), an HTML report is generated in the htmlcov folder showing each source file and which lines were executed during unit testing. Open htmlcov/index.html in a web browser to view the report. Code coverage reports help identify areas of the project that are currently not tested.

pytest and code coverage are configured in pyproject.toml.

To pass arguments to pytest through nox:

(.venv) $ nox -s test -- -k invalid_factorial

Code Style Checking

PEP 8 is the universally accepted style guide for Python code. PEP 8 code compliance is verified using Ruff. Ruff is configured in the [tool.ruff] section of pyproject.toml.

To lint code, run:

(.venv) $ nox -s lint

To automatically fix fixable lint errors, run:

(.venv) $ nox -s lint_fix

Automated Code Formatting

Ruff is used to automatically format code and group and sort imports.

To automatically format code, run:

(.venv) $ nox -s fmt

To verify code has been formatted, such as in a CI job:

(.venv) $ nox -s fmt_check

Type Checking

Type annotations allows developers to include optional static typing information to Python source code. This allows static analyzers such as mypy, PyCharm, or Pyright to check that functions are used with the correct types before runtime.

def factorial(n: int) -> int:
    ...

mypy is configured in pyproject.toml. To type check code, run:

(.venv) $ nox -s type_check

See also awesome-python-typing.

Distributing Type Annotations

PEP 561 defines how a Python package should communicate the presence of inline type annotations to static type checkers. mypy's documentation provides further examples on how to do this.

Mypy looks for the existence of a file named py.typed in the root of the installed package to indicate that inline type annotations should be checked.

Typos

Check for typos using typos

(.venv) $ nox -s typos

Continuous Integration

Continuous integration is provided by GitHub Actions. This runs all tests, lints, and type checking for every commit and pull request to the repository.

GitHub Actions is configured in .github/workflows/python.yml.

Visual Studio Code

Install the Python extension for VSCode.

Install the Ruff extension for VSCode.

Default settings are configured in .vscode/settings.json which will enable Ruff with consistent settings.

Generating Documentation

Generating a User Guide

Material for MkDocs is a powerful static site generator that combines easy-to-write Markdown, with a number of Markdown extensions that increase the power of Markdown. This makes it a great fit for user guides and other technical documentation.

The example MkDocs project included in this project is configured to allow the built documentation to be hosted at any URL or viewed offline from the file system.

To build the user guide, run,

(.venv) $ nox -s docs

and open docs/user_guide/site/index.html using a web browser.

To build the user guide, additionally validating external URLs, run:

(.venv) $ nox -s docs_check_urls

To build the user guide in a format suitable for viewing directly from the file system, run:

(.venv) $ nox -s docs_offline

To build and serve the user guide with automatic rebuilding as you change the contents, run:

(.venv) $ nox -s docs_serve

and open http://127.0.0.1:8000 in a browser.

Each time the main Git branch is updated, the .github/workflows/pages.yml GitHub Action will automatically build the user guide and publish it to GitHub Pages. This is configured in the docs_github_pages Nox session.

Generating API Documentation

This project uses mkdocstrings plugin for MkDocs, which renders Google-style docstrings into an MkDocs project. Google-style docstrings provide a good mix of easy-to-read docstrings in code as well as nicely-rendered output.

"""Computes the factorial through a recursive algorithm.

Args:
    n: A positive input value.

Raises:
    InvalidFactorialError: If n is less than 0.

Returns:
    Computed factorial.
"""

Misc

If you get a Failed to create the collection: Prompt dismissed.. error when running poetry update on Ubuntu, try setting the following environment variable:

```bash
export PYTHON_KEYRING_BACKEND=keyring.backends.null.Keyring
```

Attributions

python-blueprint for the Python package skeleton.