DEV Community

Cover image for LEVERAGING AI TO BETTER PREPARE PROJECT DOCUMENTATION FOR DEVELOPER.
nuzulfikrie
nuzulfikrie

Posted on

LEVERAGING AI TO BETTER PREPARE PROJECT DOCUMENTATION FOR DEVELOPER.

Image by leo zeng from Pixabay

Hi all,

Today, I will share on how I make use of AI to better prepare for a project. Given the flexibility nature of a software development project, having an update documentation for a developer is a must.

In the age of AI, LLM , machine learning and agentic AI , it is a waste of resources if we did not leverage on AI or in this particular scenario LLM to speed up tedious documentation process.

Generally, this is my personal approach that I wanted to use:

General flow of the process of using LLM as friend for entire developer journey

Due to space constraints, I will focus on documentation parts only.
Here are the general steps:

  1. I gather required information for the project.
  2. All relevant information is then fed to any LLM to generate documentations. I used Claude and Grok to generate documentations. Sometimes I chained the LLM output and inter-interrogate the LLM to make it produce higher quality documents.
  3. LLM then churn out base documentations, in markdowns formats consists of paragraph and flow charts - I will use mermaidJS as the go to tools as diagramming tools it is ubiquitous and generally supported by major code repository.
  4. As a rule - never trust LLM generated code or documentations so please own due diligence process and counter verify everything.
  5. All documentations are recorded inside a git repo, where all documentations are versioned accordingly.
  6. Ensure that , all documentations are verified and checked vs original spec,please engage the relevant stakeholders to ensure no document is always accurate, up to date and no errors.

Ok lets summarize and recap.

Here are the steps, simplified.

  1. obtain documentation or, rough ideas
  2. Feed to LLM , generate PRD or rough codes, provide flows and use mermaidjs as the go to tools.
  3. Counter check generated code, once LLM spit it out.
  4. All generated code are versioned in Git

All of above steps can be automated via usage of Bot or MCP.

I will share on how to automate the document generation and perhaps, make use of MCP and auto code sync in the next posting.

These are not AI generated. Thank you for reading this article, if you like it give it a like

Top comments (0)