DocAi

DocAi is a tool that allows you to automatically generate markdown-formatted documentation for your code.

Installation

npm install --save-dev docai

External LLM Mode

Prerequisites

Your code will be sent to an external provider; ensure that you have the rights to do so.

Usage Examples

Module

To document only some files:

import docai from 'docai'

await docai({
  llm: {
    apiKey: 'YOUR_API_KEY'
    modelProvider: 'mistral' | 'openAI' | 'groq'
    modelName: 'mistral-tiny' | 'gpt-4' | ...
  },
  outputDir: './generated',
  files: ['./test2.ts', './test.js']
})

To document your entire project, provide an entrypoint:

await docai({
  llm: {
    apiKey: 'YOUR_API_KEY'
    modelProvider: 'mistral' | 'openAI' | 'groq'
    modelName: 'mistral-tiny' | 'gpt-4' | ...
  },
  outputDir: './generated',
  entryPoint: './index.js'
})

To document routes from your serverless project:

await docai({
  llm: {
    apiKey: 'YOUR_API_KEY'
    modelProvider: 'mistral' | 'openAI'| 'groq'
    modelName: 'mistral-tiny' | 'gpt-4' | ...
  },
  outputDir: './generated',
  serverlessEntryPoint: './serverless.yml'
})

Optional Options

• baseDir: Defaults to the current directory. Otherwise, provide the directory path.
• llm:
  • temperature: Temperature setting for the used model (0 by default).
• deleteTmpFolder: Flag to decide whether or not to delete the temporary folder.
• tmpFolderPath: Path for the temporary folder.
• template: Prompt template send to the LLM to document the code, default value can be found here

CLI

Minimal Configuration: Run these commands at the root of your project:

API_KEY="YOUR_API_KEY" docai --output ./documentation --entrypoint ./src/index.js --modelProvider openAI --modelName gpt-3.5-turbo

or with Mistral

API_KEY="YOUR_API_KEY" docai --output ./documentation --entrypoint ./src/index.js --modelProvider mistral --modelName mistral-tiny

or with Groq

API_KEY="YOUR_API_KEY" docai --output ./documentation --entrypoint ./src/index.js --modelProvider groq --modelName mixtral-8x7b-32768

For a serverless project:

API_KEY="YOUR_API_KEY" docai --output ./documentation --modelProvider openAI --modelName gpt-3.5-turbo --serverless ./serverless.yml

Parameters:

• entrypoint: Entry point of your application. (Required)
• serverless: Path to the serverless.yml file for parsing routes in serverless projects. (Required)
• output: Destination folder path. (Required)
• modelName: Name of the LLM model to use. (Required)
• modelProvider: Name of the LLM Provider to use - openAI, mistral, groq. (Required)
• baseDir: Defaults to the current directory. Otherwise, provide the directory path. (Optional)
• temperature: Temperature setting for the used model (0 by default). (Optional)
• noDeleteTmp: Flag to decide whether or not to delete the temporary folder. (Optional)
• tmpFolderPath: Path for the temporary folder. (Optional)
• template: Prompt template send to the LLM to document the code, default value can be found here. (Optional)

Environment Variables:

Only the API_KEY environment variable is required.

Local Mode

You can run DocAi in local, with Ollama.

Please follow instructions from the Ollama documentation to install and run a model.

Usage :

Open a terminal tab, and run a model with Ollama.

Ex: ollama run llama2

Then run DocAI with CLI or Module

CLI

docai --output ./documentation --entrypoint ./src/index.js --modelName llama2 --local --baseUrl http://localhost:11434

Parameters:

• entrypoint: Entry point of your application. (Required)
• serverless: Path to the serverless.yml file for parsing routes in serverless projects. (Required)
• output: Destination folder path. (Required)
• local: Flag to local mode. (Required)
• modelName: Name of the LLM model to use. (Required)
• baseDir: Defaults to the current directory. Otherwise, provide the directory path. (Optional)
• baseUrl: Ollama REST API endpoint, default to http://localhost:11434 (Optional)

Module

import docai from 'docai'

 await docai({
   local: {
     modelName: 'llama2'
     baseUrl: 'http://localhost:11434' // default value, optional field
   },
   outputDir: './generated',
   files: ['./test2.ts', './test.js']
 })

Note: If you work in local mode, you can not pass llm object parameter.

Examples

You can examine the code found in _mock/test/raw/src and compare it to the documented version in _mock/test/generated/src.

Upcoming Features:

• Multiple entry points.
• Multilingual capabilities.
• Frontend application documentation.
• ...

Feedback and Contribution:

We value your feedback and contributions! If you encounter any issues or have suggestions for improvements, please feel free to submit an issue on our GitHub repository.