Doc-gen is an open source utility for testing the use of Artificial Intelligence (AI) in the Software Development Life Cycle (SDLC). The current version 0.1 allows the user to generate documentation from a set of source code using three different Inference Providers and their associated Large Language Models (LLMs). The Inference Providers that are supported are: OpenAI, Google AI and Hugging Face. The LLMs used can be configured in the root config.yaml file.
Doc-gen is a command line utility. It has been created and tested on Linux, however it should also run on MacOS and Windows under Windows Subsystem for Linux (WSL) with Ubuntu 22.
Doc-gen is entirely written in Python so you will need Python 3.10 or later to run this utility. You can install Python on the command line using the following commands:
$ sudo apt-get update
$ sudo apt-get install python
Doc-gen is open source and is provided as source code. In order to run it you will need to clone the doc-gen repo from GitHub. You can clone the repo with the following command:
$ git clone git@github.com:wehale/doc-gen.git
In order to do this you will need both git and an SSH key in your GitHub account. For more information on how to set this up please refer to the following documentation from GitHub.
Doc-gen supports three different Inference Providers, OpenAI, GoogleAI and Hugging Face. The cost of using these providers varies from free to pay as you go, to a monthly fee. Doc-gen is not paying for access to these providers so you will have to set up your own accounts and set the associated API Keys in your system's environment variables. For more information on doing this see below:
In order to use OpenAI as an Inference Provider you will need to set up an OpenAI account. Comprehensive information on OpenAI account setup is available here: https://platform.openai.com/docs/quickstart/account-setup. OpenAI has a free tier which is rate limited, or you can choose a pay as you go plan. More information on pricing options at OpenAI are available here: https://openai.com/api/pricing/. The pricing is not too terribly expensive (I spent $30 getting Doc-gen to the 0.1 release). But having said that, it is not free either. If free is what you are looking for you may wish to opt for one of the other Inference Providers listed below.
Once you have set up your account you will need an API Key. Information on getting an API Key for OpenAI is available here: https://platform.openai.com/api-keys
Once you have your OpenAI API Key, you will need to store it in an environment variable. Doc-gen looks for it in the OPENAI_API_KEY environment variable. For example you can add it to your environment by adding the following line to your ~/.profile file:
export OPENAI_API_KEY="sk-proj-wLMaFKEothis-is-a-fake-key-so-dont-use-it-it-is-only-an-exampleX"
Once you have added the API key to your ~/.profile file you'll need to source the file to load it into your working environment. To source the file use the following command:
$ source ~/.profile
In OpenAI, API Keys are attached to projects. You will want to create a new project and API Key for any testing you do with Doc-gen so that you have a clean sandbox in which to do your work.
In order to use Google AI as an Inference Provider you will need a Google AI (for Developers) account, which you can get through your regular Google Login if you have one for GMail. More information on Google AI for Developers is located here: https://ai.google.dev/gemini-api/docs. Use of Google AI Gemini is free ... to a point based on rate and requests per day (RPD) limits. More information on pricing is available here: https://ai.google.dev/pricing
Once you have set up your account you will need an API Key. Information on getting your API key for Google AI is located here: https://ai.google.dev/gemini-api/docs/api-key
Once you have your API Key, you'll need to install it as an environment variable. You can do this by adding it to your ~/.profile file as the following example illustrates:
export GOOGLE_API_KEY="AIz-this-is-only-an-example-dont-use-this-uQ"
Once you have added the API key to your ~/.profile file you'll need to source the file to load it into your working environment. To source the file use the following command:
$ source ~/.profile
Google AI API Keys are tied to Google Cloud projects. Pricing is also tied to the pricing model chosen for that project. As with OpenAI, if you do a lot of work with this or any other key you'll be gently pushed into pay-as-you-go pricing.
In order to use Hugging Face as an Inference Provider you will need to create a Hugging Face account. Information on creating an account is available here: https://huggingface.co/welcome. Use of Hugging Face has a (severely rate limited) free tier or you can sign up with a PRO account for what is currently $9/month. Information on pricing is available here: https://huggingface.co/pricing
As with the Inference Providers listed above, you will need an API key (called an Access Token) from Hugging Face to use the API. You can create an API Key using this documentation: https://huggingface.co/docs/hub/security-tokens
Once you have your API Key, you'll need to install it as an environment variable. You can do this by adding it to your ~/.profile file as the following example illustrates:
export HF_API_KEY="hf_sZB-this-is-only-an-example-dont-use-this-WFFG"
Once you have added the API key to your ~/.profile file you'll need to source the file to load it into your working environment. To source the file use the following command:
$ source ~/.profile
Access Tokens in Hugging Face are tied to specific use cases and can only be used for that purpose. You can create multiple Access Tokens for different use cases.
Once you have set up at least one API Key you can run doc-gen with the following command:
$ ./run.sh
The firs thing that run.sh does is create a Python virtual environment and loads all the dependencies for doc-gen into that environment. After running, run.sh automatically deactivates the virtual environment. The Python venv is stored in the dgv directory.
Doc-gen runs according to the configuration settings in the config.yaml file. Complete information on the settings in the config.yaml file are located in the file itself. config.yaml
Doc-gen is written to only run against Inference Providers with configured API Keys, If the settings in config.yaml are set to run against an Inference Provider for which no key is found, doc-gen will not run generation for that IP.
Doc-gen uses Python threading to run all provided files against all Inference Providers at the same time.
Currently the output of Doc-gen is a set of markdown .md files and an HTML site generated from those .md files through Sphinx. All of the output is generated into the build directory.
By default, doc-gen runs against the two code files in the ./samples directory calc.c and SpellChecker.java. If you would like to run it against some other code, you can either add the code files of your choosing into the ./samples director OR change the input.doc_path setting in the config.yaml configuration file.
The last thing that run.sh does is to give the user the option to open the created HTML in the default browser of their choosing. This way the generated content can be viewed side by side.
For an example of what doc-gen's html output looks like please see the published output here: https://wehale.github.io/doc-gen-pub/latest/