Using The Lad (AI Assistant)

​

Overview

​

Basic Usage

applad instruct "add a posts table with title, body, and author_id"

1. Analyzes the instruction
2. Creates database/tables/posts.yaml with appropriate fields
3. Generates a migration file
4. Shows what it will do (always --dry-run first)
5. Asks for confirmation
6. Applies changes
7. Records the prompt in the audit trail

​

Configuration

ai:
  enabled: true
  provider: "anthropic"  # anthropic | openai | google | ollama
  model: "claude-sonnet-4-6"

  features:
    scaffolding: true
    debugging: true
    monitoring: true
    migrations: true
    infrastructure: true
    security: true
    dry_run: true

​

Provider Configuration

applad secrets set ANTHROPIC_API_KEY
applad secrets set OPENAI_API_KEY

ai:
  enabled: true
  provider: "openai"
  model: "gpt-4-turbo"

​

Common Tasks

​

Create Tables

applad instruct "add a users table with email, name, and role"

• database/tables/users.yaml with fields, indexes, and permissions
• Migration file
• Updates .env.example if new variables are needed

​

Add Fields to Existing Tables

applad instruct "add avatar_url field to users table"

• Migration to add the field
• Updates table YAML definition

​

Create Functions

applad instruct "create a function to send welcome emails when users sign up"

• functions/send-welcome-email.yaml with event trigger on auth.user.created
• Boilerplate function code in your preferred runtime
• Shows where to place the code

​

Configure Storage

applad instruct "add a public avatars bucket that accepts images up to 5mb"

• storage/buckets/avatars.yaml with appropriate configuration
• Permission rules

​

Set Up Deployments

applad instruct "create a web deployment for myapp.com using the main branch"

• deployments/web.yaml with domain, source, and build configuration
• Security headers
• SSL configuration

​

Add Feature Flags

applad instruct "create a feature flag for the new dashboard"

• flags/new-dashboard.yaml with boolean type
• Per-environment defaults

​

Advanced Usage

​

Infrastructure Changes

applad instruct "add a staging environment using a VPS at staging.myapp.com"

• project.yaml with new environment definition
• Infrastructure target configuration
• Environment-specific overrides

​

Security Audits

applad instruct "audit the posts table permissions for security issues"

• Permission rules
• Filter expressions
• Potential security gaps
• Suggests improvements

​

Debugging

applad instruct "why is the send-welcome-email function failing?"

• Recent function logs
• Configuration
• Environment variables
• Common failure patterns
• Suggests fixes

​

Migration Assistance

applad instruct "create a migration to add fulltext search to posts"

• Migration SQL
• Down migration for rollback
• Updates table YAML with fulltext index

​

Dry-Run Mode

applad instruct --dry-run "add a comments table"

• Which files would be created or modified
• Content of each file
• Environment variables that would be added
• Migrations that would be generated

​

Access Control

applad instruct "apply this change to production"

ERROR Instruction requires infrastructure:apply:production scope
  Your key (SHA256:def456... bob@workstation) does not have this scope.
  Ask an admin to grant it with:
    applad access grant bob@acme-corp \
      --scope "infrastructure:apply:production" \
      --project mobile-app

​

Audit Trail

applad security events list --type instruct_action

• Exact prompt
• SSH key identity
• Files created or modified
• When it happened
• Success or failure

​

Multi-Step Instructions

applad instruct "create a blog feature with posts, comments, and categories tables, plus a function to send notifications when someone comments"

1. Creates three table definitions
2. Defines relations between tables
3. Generates migrations
4. Creates the notification function
5. Sets up event trigger
6. Shows the complete plan
7. Asks for confirmation

​

Best Practices

applad instruct --dry-run "migrate production to new schema"

applad instruct "add posts table"
# Review and apply

applad instruct "add comments table with relation to posts"
# Review and apply

applad security events list --type instruct_action --since today

​

Example Workflows

​

Building a Feature

# 1. Create tables
applad instruct "add a posts table with title, body, author_id, and status enum"

# 2. Add storage
applad instruct "create a post-images bucket for uploaded post images"

# 3. Create function
applad instruct "create a function to process post images on upload"

# 4. Add feature flag
applad instruct "create a feature flag for the new posts feature"

# 5. Set up deployment
applad instruct "add the posts feature to the web deployment"

​

Debugging Production

# Check what's wrong
applad instruct "why are users unable to upload avatars?"

# Get logs
applad instruct "show recent errors from the upload-avatar function"

# Analyze config
applad instruct "check if the avatars bucket permissions are correct"

​

Infrastructure Changes

# Add environment
applad instruct "add a staging environment on staging.myapp.com"

# Configure cloud resources
applad instruct "add S3 storage for production using eu-west-1"

# Update deployment
applad instruct "configure the web deployment to build on every push to main"

​

Limitations

• Execute arbitrary code
• Access external systems without explicit configuration
• Override SSH key permission scopes
• Make changes without human confirmation (unless explicitly automated)
• Perform destructive operations without --dry-run review

​

Disabling AI Features

ai:
  enabled: false

ai:
  enabled: true
  provider: "anthropic"
  model: "claude-sonnet-4-6"

  features:
    scaffolding: true
    debugging: true
    monitoring: true
    migrations: false      # Disable migration generation
    infrastructure: false  # Disable infrastructure changes
    security: true
    dry_run: true

​

Privacy and Data

• Your config structure is sent to the AI provider
• Actual secret values are never sent (only ${VAR} references)
• Prompts and responses are logged in your audit trail
• You can use self-hosted models (Ollama) for full privacy

​

Self-Hosted AI

ai:
  enabled: true
  provider: "ollama"
  model: "llama2"
  config:
    endpoint: "http://localhost:11434"

​

Next Steps