Experts Setup

Overview

The Experts feature in HAI Build enables context-aware code generation tailored to specific technologies, frameworks, or domain standards. You can use built-in experts or create custom ones with your own guidelines and reference documentation.

Experts are currently an upcoming feature in HAI Build. Check the latest release notes for availability.

How Experts Work

Experts provide:

Technology-specific guidelines and best practices
Reference documentation for accurate code generation
Consistent coding standards across your team
Domain-specific context for AI-powered development

When you select an expert, HAI Build incorporates their guidelines into the AI’s context, ensuring generated code follows your standards.

Built-in Experts

HAI Build includes four built-in experts available in read-only mode:

.NET

Best practices for C# and .NET development

Terraform

Infrastructure as Code guidelines and patterns

Node.js

JavaScript/TypeScript server-side development

Go

Go language idioms and conventions

Built-in experts provide predefined guidelines for quick reference and cannot be modified.

Custom Experts

Create custom experts to define your own coding standards, architectural patterns, or domain-specific requirements.

Expert Structure

Custom experts are stored in the .hai-experts directory at your workspace root:

my-project/
├── .hai-experts/
│   ├── my-expert/
│   │   ├── metadata.json
│   │   ├── prompt.md
│   │   └── docs/
│   │       ├── document-1.md
│   │       ├── document-2.md
│   │       └── status.json
│   └── another-expert/
│       ├── metadata.json
│       └── prompt.md
├── .git/
└── src/

File Structure

metadata.json

Stores expert metadata and configuration:

{
  "name": "My Custom Expert",
  "createdAt": "2024-03-03T12:00:00Z",
  "documentLinks": [
    "https://example.com/docs/api-guide",
    "https://example.com/docs/best-practices",
    "https://example.com/docs/architecture"
  ]
}

string

required

Display name of the expert

string

required

ISO 8601 timestamp of creation

array

default:"[]"

Up to 3 reference document URLs (optional)

prompt.md

Contains the expert’s guidelines and instructions (required):

# Custom Expert Guidelines

## Code Style
- Use functional components with hooks
- Prefer TypeScript strict mode
- Follow Airbnb style guide

## Architecture
- Use repository pattern for data access
- Implement dependency injection
- Separate concerns with clean architecture

## Testing
- Write unit tests for all business logic
- Use Jest and React Testing Library
- Aim for 80% code coverage

docs/ (optional)

Processed reference documents:

Markdown files for each document link
status.json tracks processing status

// status.json
{
  "processed": 2,
  "total": 3,
  "documents": [
    {
      "url": "https://example.com/docs/api-guide",
      "status": "completed",
      "file": "document-1.md"
    },
    {
      "url": "https://example.com/docs/best-practices",
      "status": "completed",
      "file": "document-2.md"
    },
    {
      "url": "https://example.com/docs/architecture",
      "status": "pending",
      "file": null
    }
  ]
}

Creating a Custom Expert

Open Experts Interface

Navigate to the Experts section in HAI Build sidebar

Click Create Expert

Select “Create Custom Expert” to start the wizard

Enter Expert Details

Provide required information:

string

required

Descriptive name for your expert (e.g., “React Best Practices”)

markdown

required

Coding standards, patterns, and best practices in Markdown format

array

default:"[]"

Up to 3 reference document URLs (optional)

Save Expert

HAI Build creates the expert folder structure in .hai-experts/

Process Documents (if provided)

HAI Build fetches and processes reference documents in the background

Example: React Expert

Create a React expert with your team’s standards: Expert Name: React Best Practices Guidelines (prompt.md):

# React Development Standards

## Component Structure
- Use functional components with TypeScript
- Implement custom hooks for reusable logic
- Keep components under 200 lines

## State Management
- Use React Context for global state
- Implement Redux for complex applications
- Avoid prop drilling beyond 2 levels

## Styling
- Use CSS Modules or styled-components
- Follow BEM naming convention
- Implement responsive design with mobile-first approach

## Performance
- Memoize expensive computations with useMemo
- Use React.memo for pure components
- Implement code splitting with React.lazy

## Testing
- Write integration tests with Testing Library
- Test user interactions, not implementation
- Aim for 80% code coverage

Document Links:

https://react.dev/reference/react
https://testing-library.com/docs/react-testing-library/intro
https://styled-components.com/docs

Example: API Integration Expert

Create an expert for your API integration standards: Expert Name: API Integration Standards Guidelines (prompt.md):

# API Integration Standards

## HTTP Client
- Use axios with interceptors for auth
- Implement retry logic with exponential backoff
- Handle errors with custom error classes

## Request Patterns
- Separate API calls into service layers
- Use TypeScript interfaces for request/response
- Implement request cancellation for search

## Authentication
- Store tokens in httpOnly cookies
- Implement token refresh mechanism
- Handle 401 responses globally

## Error Handling
- Show user-friendly error messages
- Log errors to monitoring service
- Implement fallback UI for failures

## Caching
- Use React Query for data fetching
- Implement stale-while-revalidate strategy
- Cache static data in localStorage

Document Links:

https://axios-http.com/docs/intro
https://tanstack.com/query/latest/docs/react/overview

Using Experts

Once created, experts can be selected when starting a new task or conversation with HAI Build.

Select Expert

Choose the relevant expert from the dropdown when starting a task

Expert Context Applied

HAI Build incorporates the expert’s guidelines into the AI’s context

Generate Code

Code generation follows the expert’s standards and best practices

For consistent results across your team, commit .hai-experts/ to your repository (excluding any sensitive data).

Managing Experts

Editing Experts

To modify a custom expert:

Navigate to .hai-experts/expert-name/
Edit prompt.md or metadata.json
Save changes
Restart HAI Build or reload the expert

Deleting Experts

To remove a custom expert:

Delete the expert folder from .hai-experts/
Refresh HAI Build

Built-in experts cannot be modified or deleted.

Share experts with your team by:

Committing .hai-experts/ to version control
Excluding sensitive information from metadata.json
Documenting expert usage in your team’s documentation

Reference Document Processing

When you provide document links, HAI Build:

Fetches the content from the URLs
Converts to Markdown format
Stores in the docs/ folder
Updates status.json with processing status

Document Limits

Maximum 3 document links per expert
Documents must be publicly accessible
Supported formats: HTML, Markdown, PDF (depending on implementation)

Status Tracking

Monitor document processing in status.json:

{
  "documents": [
    {
      "url": "https://example.com/guide",
      "status": "completed",
      "file": "document-1.md"
    },
    {
      "url": "https://example.com/api",
      "status": "failed",
      "error": "URL not accessible"
    }
  ]
}

Status values:

pending - Processing not started
in_progress - Currently fetching/converting
completed - Successfully processed
failed - Error during processing

Best Practices

Be Specific

Provide clear, actionable guidelines rather than vague suggestions

Use Examples

Include code examples in guidelines to demonstrate patterns

Keep Updated

Regularly review and update expert guidelines as standards evolve

Version Control

Commit experts to git for team consistency and version history

Focus on Domain

Create experts for specific domains rather than general coding

Reference Quality Docs

Link to official documentation and authoritative sources

Troubleshooting

Expert Not Appearing

Check:

Expert folder exists in .hai-experts/
metadata.json is valid JSON
prompt.md exists and is not empty
Restart VS Code or reload HAI Build

Document Processing Failed

Solutions:

Verify URLs are publicly accessible
Check status.json for error messages
Try removing and re-adding document links
Ensure network connectivity

Expert Context Not Applied

Solutions:

Verify expert is selected before starting task
Check expert guidelines are clear and specific
Ensure expert folder structure is correct
Try recreating the expert

Cannot Modify Built-in Expert

Built-in experts (.NET, Terraform, Node.js, Go) are read-only by design.Solution: Create a custom expert based on the built-in one:

Copy guidelines from built-in expert
Create new custom expert
Modify as needed

Example Experts

Python Data Science Expert

Guidelines:

# Python Data Science Standards

## Libraries
- Use pandas for data manipulation
- NumPy for numerical operations
- matplotlib/seaborn for visualization
- scikit-learn for ML models

## Code Style
- Follow PEP 8 style guide
- Use type hints for function signatures
- Document with NumPy-style docstrings

## Data Processing
- Validate data with pandas-schema
- Handle missing values explicitly
- Use vectorized operations over loops

## Model Development
- Split data: 70% train, 15% validation, 15% test
- Use cross-validation for hyperparameter tuning
- Track experiments with MLflow

DevOps Expert

Guidelines:

# DevOps and Infrastructure Standards

## CI/CD
- Use GitHub Actions for automation
- Implement automated testing in pipeline
- Deploy to staging before production

## Infrastructure
- Define infrastructure as code with Terraform
- Use modules for reusable components
- Store state in remote backend

## Containerization
- Use multi-stage Docker builds
- Scan images for vulnerabilities
- Keep images under 500MB

## Monitoring
- Implement health checks for all services
- Set up alerts for critical metrics
- Use structured logging

Mobile Development Expert

Guidelines:

# Mobile Development Standards

## Architecture
- Use MVVM architecture pattern
- Implement dependency injection
- Separate UI, business logic, and data layers

## State Management
- Use Redux/MobX for complex state
- Implement offline-first approach
- Sync data in background

## Performance
- Lazy load images and heavy components
- Implement pagination for lists
- Optimize bundle size

## Testing
- Write unit tests for business logic
- Implement E2E tests with Detox
- Test on both iOS and Android

Next Steps

Settings

Configure extension settings

LLM Providers

Set up AI model providers

Telemetry

Configure monitoring and analytics

Get Started

Core Features

Configuration

Usage Guides

Advanced

Overview

How Experts Work

Built-in Experts

.NET

Terraform

Node.js

Go

Custom Experts

Expert Structure

File Structure

Creating a Custom Expert

Example: React Expert

Example: API Integration Expert

Using Experts

Managing Experts

Editing Experts

Deleting Experts

Reference Document Processing

Document Limits

Status Tracking

Best Practices

Be Specific

Use Examples

Keep Updated

Version Control

Focus on Domain

Reference Quality Docs

Troubleshooting

Example Experts

Next Steps

Settings

LLM Providers

Telemetry

​Overview

​How Experts Work

​Built-in Experts

.NET

Terraform

Node.js

Go

​Custom Experts

​Expert Structure

​File Structure

​Creating a Custom Expert

​Example: React Expert

​Example: API Integration Expert

​Using Experts

​Managing Experts

​Editing Experts

​Deleting Experts

​Sharing Experts

​Reference Document Processing

​Document Limits

​Status Tracking

​Best Practices

Be Specific

Use Examples

Keep Updated

Version Control

Focus on Domain

Reference Quality Docs

​Troubleshooting

​Example Experts

​Next Steps

Settings

LLM Providers

Telemetry

Overview

How Experts Work

Built-in Experts

Custom Experts

Expert Structure

File Structure

Creating a Custom Expert

Example: React Expert

Example: API Integration Expert

Using Experts

Managing Experts

Editing Experts

Deleting Experts

Sharing Experts

Reference Document Processing

Document Limits

Status Tracking

Best Practices

Troubleshooting

Example Experts

Next Steps