Learning Commands

Commands for the Aura Frog Learning System.

---

Overview

The learning system enables Aura Frog to improve over time by:

• Automatic: Collecting feedback, metrics, agent usage via hooks
• Manual: Analyzing patterns and applying improvements via commands

Full Guide: docs/LEARNING_SYSTEM.md

---

Commands

Command	Purpose	When to Use
/learn:setup	Create database schema	First time setup
/learn:status	Show status and stats	Check if working
/learn:feedback	Submit manual feedback	Report issues or successes
/learn:analyze	Generate insights	Weekly or on-demand
/learn:apply	Apply improvements	After analysis

---

Command Details

/learn:setup

Sets up the Supabase database schema.

Prerequisites:

1. Run bootstrap.sql in Supabase SQL Editor first
2. Set SUPABASE_URL and SUPABASE_SECRET_KEY in .envrc

What it does:

1. Checks environment variables
2. Verifies bootstrap function exists
3. Creates all tables, views, functions
4. Confirms success

---

/learn:status

Shows learning system status and statistics.

What it shows:

• Connection status
• Feedback count
• Workflow metrics count
• Agent success rates
• Pending improvements

---

/learn:feedback

Manually submit feedback to the learning system.

Feedback types:

• Success - Report what worked well
• Correction - Report when you had to fix something
• Agent Issue - Report wrong agent selection
• Workflow Issue - Report workflow problems
• Suggestion - General improvement ideas

Options:

• --type <type> - Skip type selection (success, correction, agent-issue, workflow-issue, suggestion)
• --message <msg> - Quick feedback message
• --severity <1-5> - Issue severity (for corrections)

Examples:

/learn:feedback                                    # Interactive
/learn:feedback --type success --message "Great!" # Quick positive
/learn:feedback --type correction --severity 3    # Report fix

---

/learn:analyze

Analyzes collected data and generates insights.

What it does:

1. Fetches data from Supabase views
2. Identifies success/failure patterns
3. Generates optimization suggestions
4. Saves new patterns to database
5. Shows analysis report

Options:

• --focus agents - Focus on agent performance
• --focus workflows - Focus on workflow patterns
• --focus feedback - Focus on user corrections

---

/learn:apply

Reviews and applies learned improvements.

What it does:

1. Fetches pending suggestions
2. Shows each with proposed change
3. Applies if user approves
4. Creates backup before changes
5. Marks as applied in database

Options:

• --auto - Auto-apply high confidence (≥80%)
• --preview - Show without applying
• --id <id> - Apply specific suggestion

---

Quick Start

# 1. First time setup
/learn:setup

# 2. Verify it's working
/learn:status

# 3. After using Aura Frog for a while...
/learn:analyze

# 4. Apply improvements
/learn:apply

---

Auto vs Manual

What	How	When
Data collection	Automatic (hooks)	Every session
Send to Supabase	Automatic (on session end)	Every session
Manual feedback	Manual (/learn:feedback)	Anytime (optional)
Analysis	Manual (/learn:analyze)	Weekly recommended
Apply changes	Manual (/learn:apply)	After analysis

---

Configuration

Required in .envrc:

export SUPABASE_URL="https://your-project.supabase.co"
export SUPABASE_SECRET_KEY="your-secret-key"
export AF_LEARNING_ENABLED="true"

---

Troubleshooting

Issue	Solution
"Not configured"	Set env vars, run source .envrc
"Bootstrap not found"	Run bootstrap.sql in Supabase SQL Editor
"Table not found"	Run /learn:setup
Empty results	Use Aura Frog more to collect data

---

Version: 1.0.0