OPEA Documentation Generation

These instructions walk you through generating the OPEA project documentation and publishing it to https://opea-project.github.io. You can also use these instructions to generate the OPEA documentation on your local system.

Documentation Overview

OPEA project content is written using combination of markdown (.md) and reStructuredText (.rst) markup languages (with Sphinx extensions), and processed using Sphinx to create a formatted stand-alone website. Developers can view this content either in its raw form as .rst markup files, or you can generate the HTML content and view it with a web browser directly on your workstation. The best reading experience is by viewing the generated HTML at https://opea-project.github.io.

You can read details about markdown, reStructuredText, and Sphinx from their respective websites.

The project’s documentation contains the following items:

• ReStructuredText and markdown source files used to generate documentation found at the https://opea-project.github.io website. All of the documentation sources are found in the github.com/opea-project repos, rooted in the docs repo. There’s also documentation in the repos where the project’s code is maintained: GenAIComps, GenAIEval, GenAIExamples, and GenAIInfra.

Figure 3 Documentation Generation Flow

Set Up the Documentation Working Folders

You’ll need git installed to get the working folders set up:

• For an Ubuntu development system use:

  sudo apt install git
  

Here’s the recommended folder setup for documentation contributions and generation, a parent folder called opea-project holds six locally cloned repos from the opea-project. You can use a different name for the parent folder but the doc build process assumes the repo names are as shown here:

opea-project
├── docs
├── GenAIComps
├── GenAIEval
├── GenAIExamples
├── GenAIInfra
├── opea-project.github.io

The parent opea-project folder is there to organize the cloned repos from the project. If you have repo publishing rights, we’ll also be cloning the publishing repo opea-project.github.io later in these steps.

In the following steps, you’ll create a fork of all the upstream OPEA project repos needed to build the documentation to your personal GitHub account, clone your personal fork to your local development computer, and then link that to the upstream repo as well. You’ll only need to do this once to set up the folder structure:

1. Use your browser to visit https://github.com/opea-project and do a fork of the docs repo to your personal GitHub account.)

   

2. At a command prompt, create a working folder on your development computer and clone your personal docs repository:

   cd ~
   mkdir opea-project && cd opea-project
   git clone https://github.com/<github-username>/docs.git
   

3. For the cloned local repo, tell git about the upstream repo:

   cd docs
   git remote add upstream https://github.com/opea-project/docs.git
   

   After that, you’ll have origin pointing to your cloned personal repo and upstream pointing to the project repo.

4. Do the same steps (fork to your personal account, clone to your local computer, and setup the git upstream remote) for the other repos containing project documentation, replacing the docs.git repo name in the previous step with the appropriate repo name in this list:

   • GenAIComps

   • GenAIEval

   • GenAIExamples

   • GenAIInfra

5. If you haven’t done so already, be sure to configure git with your name and email address for the Signed-off-by line in your commit messages:

   git config --global user.name "David Developer"
   git config --global user.email "david.developer@company.com"
   

Install the Documentation Tools

Our documentation processing has been tested to run with Python 3.8.10 and later, and these other tools:

• sphinx version: 7.3.0

• docutils version: 0.20

• sphinx-rtd-theme version: 2.0.0

• sphinx-tabs version: 3.4.5

• myst-parser version: 3.0.1

• sphinxcontrib-mermaid version: 0.9.2

• pymarkdownlnt version: 0.9.21

Depending on your Linux version, install the needed tools. You should consider using the Python virtual environment tools to maintain your Python environment from being changed by other work on your computer.

For Ubuntu, use:

sudo apt install python3-pip python3-wheel make graphviz

Then use pip3 to install the remaining Python-based tools specified in the scripts/requirements.txt file

cd ~/opea-project/docs
pip3 install --user -r scripts/requirements.txt

Use this command to add $HOME/.local/bin to the front of your PATH so the system will find expected versions of these Python utilities such as sphinx-build (you should first check whether this folder is already on your path):

printf "\nexport PATH=\$HOME/.local/bin:\$PATH" >> ~/.bashrc

Important

You will need to open a new terminal for this change to take effect. Adding this to your ~/.bashrc file ensures it is set by default.

And with that you’re ready to generate the documentation.

Note

We’ve provided a script in the docs repo you can run to show what versions of the documentation building tools are installed and compare with the tool versions shown above. This tool will also verify you’re using tool versions known to work together:

docs/scripts/show-versions.py

for example:

~/opea-project/docs$ scripts/show-versions.py

doc build tool versions found on your system per /home/david/opea-project/docs/scripts/requirements.txt...

sphinx                    version: 7.3.0
docutils                  version: 0.20
sphinx-rtd-theme          version: 2.0.0
sphinx-tabs               version: 3.4.5
myst-parser               version: 3.0.1
sphinx-md                 version: 0.0.3
sphinxcontrib-mermaid     version: 0.9.2
pymarkdownlnt             version: 0.9.21

Documentation Presentation Theme

Sphinx supports easy customization of the generated HTML documentation appearance through the use of themes. The sphinx-rtd-theme (Read The Docs) theme is installed as part of the requirements.txt list above. Tweaks to the standard read-the-docs appearance are added by using CSS and JavaScript customization found in doc/_static, and theme template overrides found in doc/_templates. If you change to another theme, you’ll need to tweak these customizations, not something for the faint of heart.

The Sphinx build system creates document cache information that attempts to expedite documentation rebuilds, but occasionally can cause an unexpected error or warning to be generated. Doing a make clean to create a clean generation environment and a make html again generally fixes these issues.

Run the Documentation Processors

The docs folder (with all cloned sibling repos) have all the doc source files, images, extra tools, and Makefile for generating a local copy of the OPEA technical documentation. It’s best to start with a clean doc-build environment so use make clean to remove the _build working folder if it exists. The Makefile creates the _build folder (if it doesn’t exist) and copies all needed files from these cloned repos into the _build/rst working folder.

cd ~/opea-project/docs
make clean
make html

Depending on your development system, it will take less a minute to collect and generate the HTML content. When done, you can view the HTML output in ~/opea-project/docs/_build/html/index.html.

As a convenience, there’s a make target that will cd to the _build/html folder and run a local Python web server:

make server

and use your web browser to open the URL: http://localhost:8000. When done, press ctrl-C in your command-prompt window to stop the web server.

Publish Content

If you have merge rights to the opea-project repo called opea-project.github.io, you can update the public project documentation found at https://opea-project.github.io.

You’ll need to do a one-time clone of the upstream repo (we publish directly to the upstream repo rather than to a personal forked copy):

cd ~/opea-project
git clone https://github.com/opea-project/opea-project.github.io.git

Then, after you’ve verified the generated HTML produced by make html looks good, you can push to the publishing site with:

make publish

This uses git commands to synchronize the new content with what’s already published and will delete files in the publishing repo’s latest folder that are no longer needed. New or changed files from the newly-generated HTML content are pushed to the GitHub pages publishing repo. The public site at https://opea-project.github.io will be automatically updated by the GitHub pages system, typically within a few minutes.

Document Versioning

The https://opea-project.github.io site has a document version selector at the top of the left nav panel. The contents of this version selector are defined in the conf.py sphinx configuration file, specifically something like this:

html_context = {
   'current_version': current_version,
   'docs_title': docs_title,
   'is_release': is_release,
   'versions': ( ("latest", "/latest/"),
                 ("0.8", "/0.8/"),
                 ("0.7", "/0.7/"),
               )
    }

As new versions of OPEA documentation are added, typically when a new release is made, update this versions selection list to include the version number and publishing folder. Note that there’s no direct selection to go to a newer version from an older one, without going to latest first.

By default, documentation build and publishing both assume we’re generating documentation for the main branch and publishing to the /latest/ area on https://opea-project.github.io. When we’re generating the documentation for a tagged version (e.g., 0.8), check out that version of all the component repos, and add some extra flags to the make commands:

version=0.8
for d in docs GenAIComps GenAIExamples GenAIEval GenAIInfra ; do
 cd ~/opea-project/$d
 git checkout $version
done

cd ~/opea-project/docs
make clean
make DOC_TAG=release RELEASE=$version html
make DOC_TAG=release RELEASE=$version publish

Filter Expected Warnings

Alas, there are some known issues with the Sphinx processing that generate warnings. We’ve added a post-processing filter on the output of the documentation build process to check for “expected” warning messages in the generated log output. By doing this, only “unexpected” messages will be reported and cause the build process to fail with a message:

New errors/warnings found, please fix them:

followed by messages that weren’t expected. Note that the file names shown in the error/warning messages will be for files in the _build/rst folder (copied from the repos). For example,

New errors/warnings found, please fix them:
==============================================

/home/david/opea-project/docs/_build/rst/GenAIInfra/kubernetes-addons/Observability/README.md:5: WARNING: Non-consecutive header level increase; H1 to H4 [myst.header]
/home/david/opea-project/docs/_build/rst/GenAIInfra/kubernetes-addons/Observability/README.md:111: WARNING: Non-consecutive header level increase; H3 to H6 [myst.header]

For files copied from repos other than the docs repo, you’ll see the repo name in the file path, for example, _build/rst/GenAIInfra with the path to specific file with an issue. For example, the warnings shown here indicate a heading level problem on lines 5 and 111 in GenAIInfra/kubernetes-addons/Observability/README.md.

If you do a make html without first doing a make clean, there may be files left behind from a previous build that can cause some unexpected messages to be reported.

If all messages were filtered away, the build process will report as successful, reporting:

No new errors/warnings.

The output from the Sphinx build is processed by the Python script scripts/filter-known-issues.py together with a set of filter configuration files in the .known-issues folder. (This filtering is done as part of the Makefile.)

You can modify the filtering by adding or editing a conf file in the .known-issues folder, following the examples found there.