🤖 LLM-Assisted Development Guide

A systematic approach to building applications with AI assistance

📚 Table of Contents

• Getting Started
• Development Process
• Best Practices
• Documentation Strategy
• Tips & Troubleshooting
• Coming Soon

Getting Started

Before diving into development, it's crucial to:

1. Clearly define your application concept
2. Map out core functionality
3. Plan the user experience (UX)
4. List and prioritize key features

Initial Planning Prompt

Use this template to kickstart your project:

Help me define and spec out a {insert language i.e. Python} based application using {insert framework i.e. Streamlit}.

Objective: Create a plan that we can use with an LLM-assisted coding agent/IDE to generate code based on our prompts.

Please include:
1. Application concept and core functionality
2. UX/UI requirements and user flow
3. Key features and priority order
4. Required libraries and technical stack
5. Data inputs and user stories

This prompt will generate a boilerplate plan that you can refine according to your specific requirements.

Development Process

The LLM-assisted development process follows these key steps:

1. Plan application and define requirements
2. Generate focused development prompts
3. Implement features incrementally
4. Update documentation after each change
5. Use documentation as context for next steps

1. Creating Development Prompts

Use this template to generate focused development prompts:

"I have a plan for a {framework} app here: @plan.md
Project scaffolding is described in: @README.md

Please write prompts to build each part of the app, ensure each prompt includes:

1. Implementation Requirements:
   - Feature specifications
   - Error handling approach
   - Testing procedures
   - Documentation updates

2. Project Context:
   - Project: {project_name}
   - Description: {project_description}
   - Documentation Paths:
     README: {readme_path}
     CHANGELOG: {changelog_path}
     BUGS: {bugs_path}

3. Development Guidelines:
   - Preserve existing functionality
   - Maintain UI/UX consistency
   - Implement features incrementally
   - Update documentation continuously"

2. Implementation Example

Here's a structured prompt for implementing specific features:

Project Context:
You are assisting with the Report Insights App, a Streamlit-based application for analyzing ad server reports.

Documentation References:
- README: @README.md 
- CHANGELOG: @CHANGELOG.md 
- BUGS: @BUGS.md
- Framework Docs: @docs/streamlit-docs 

Development Guidelines:
- Preserve existing functionality
- Maintain consistent UI/UX
- Implement features incrementally
- Include comprehensive error handling
- Update documentation after each change
- Add appropriate tests

Feature Implementation:
Implement file upload functionality:
1. Create file uploader (XLSX/CSV support)
2. Add format validation
3. Implement data preview

Technical Requirements:
- Use st.file_uploader with type restrictions
- Show file metadata (name, size, type)
- Display first 5 rows of data
- Handle invalid files gracefully

Testing Requirements:
- Valid XLSX/CSV files
- Invalid file formats
- Error message verification

3. Documentation Updates

After each feature implementation:

1. Update README.md with:
   - New functionality
   - Updated project structure
   - Usage instructions

2. Document in BUGS.md:
   - Known issues
   - Workarounds
   - Future improvements

3. Update CHANGELOG.md:
   - Feature additions
   - Bug fixes
   - Breaking changes

Best Practices

1. Context Management

• Define frameworks and libraries upfront
• Keep prompts focused and specific
• Use documentation references instead of copying code
• Break complex features into smaller tasks

2. Documentation Strategy

• Maintain comprehensive documentation
• Update docs after each feature implementation
• Include project structure and dependencies
• Document known issues and solutions

3. Error Prevention

• Validate LLM outputs against requirements
• Test incrementally
• Keep context window focused
• Use clear, specific prompts

Tips & Troubleshooting

Recommended Tools

• jina.ai - Convert a single URL to an LLM-friendly input (useful for single documentation pages)
• firecrawl.dev - Crawl multiple URLs and convert page content to LLM-friendly input (useful for multiple documentation pages)
• llms.txt - A standardised way to provide website information to LLMs, similar to robots.txt for search engines (provide another useful source of LLM-friendly documentation

Common Issues

1. Context Overflow

   • Break down prompts into smaller chunks
   • Reference documentation instead of copying

2. Inconsistent Output

   • Provide more specific requirements
   • Include example patterns
   • Use incremental development

3. Feature Regression

   • Maintain clear documentation
   • Use documentation as context
   • Implement changes incrementally

Coming Soon

• Unit testing
• Version control

📝 Pro Tip: Always include project structure, component details, and other relevant information in your documentation files. Use these as context for subsequent development prompts to maintain consistency and prevent unintended code changes.

License

This project is licensed under the MIT License - see the LICENSE file for details.