ai-flow-tester

name	ai-flow-tester
description	Test and validate Genkit AI flows with sample inputs
disable-model-invocation	false
argument-hint	[flow-name] [--input FILE] [--compare-output] [--prompt PROMPT]
allowed-tools	Bash, Read, Write, Edit, Glob

AI Flow Tester

Test and validate Genkit AI flows for the Planar Nexus deck coach and opponent generation systems.

Usage

Test a specific flow

/ai-flow-tester ai-deck-coach-review

Test with custom input file

/ai-flow-tester ai-opponent-deck-generation --input test-deck.json

Test with custom prompt

/ai-flow-tester ai-deck-coach-review --prompt "Analyze this Modern deck: 4x Ragavan, 4x Lightning Bolt..."

Compare output against expected schema

/ai-flow-tester ai-deck-coach-review --compare-output

Available Flows

Located in /src/ai/flows/:

1. ai-deck-coach-review - Analyzes decklists and generates strategic suggestions

   • Input: Decklist with card names and quantities
   • Output: Strategic analysis, suggestions, weakness identification

2. ai-opponent-deck-generation - Creates AI opponent decks

   • Input: Format, difficulty level, archetype preferences
   • Output: Generated decklist matching criteria

Workflow

1. Flow Discovery

• Lists all available flows in /src/ai/flows/
• Reads flow configuration to understand input/output schemas

2. Test Execution

• Runs the Genkit flow with test input
• Captures output and validates against Zod schema
• Reports any errors or validation failures

3. Output Analysis

• Displays structured output
• Compares against expected schema (if --compare-output)
• Highlights any missing or malformed fields

4. Iteration

• Allows prompt refinement
• Re-runs flow with adjusted parameters
• Tracks improvements across iterations

Test Input Templates

Deck Coach Review Test Input

{
  "decklist": "4x Ragavan, Nimble Pilgrim\n4x Lightning Bolt\n4x Murktide Regent\n...",
  "format": "Modern",
  "focus": "overall"
}

Opponent Deck Generation Test Input

{
  "format": "Modern",
  "difficulty": "intermediate",
  "archetype": "aggro",
  "colorIdentity": "red"
}

Validation Checks

• ✅ Flow executes without errors
• ✅ Output matches Zod schema
• ✅ Response time < 10 seconds
• ✅ Structured output is valid JSON
• ✅ All required fields present

Error Handling

• Flow not found: Lists available flows and suggests closest match
• Schema validation error: Shows which fields failed and why
• Timeout: Retries with exponential backoff (max 3 attempts)
• API rate limit: Waits and retries after delay

Examples

Quick smoke test

/ai-flow-tester ai-deck-coach-review --prompt "4x Lightning Bolt, 4x Lava Spike, 20x Mountain"

Full validation with schema check

/ai-flow-tester ai-opponent-deck-generation --input src/ai/flows/test-inputs/opponent-deck.json --compare-output

Iterative prompt tuning

/ai-flow-tester ai-deck-coach-review --prompt "Analyze this Legacy Storm deck..."
# Review output, then refine:
/ai-flow-tester ai-deck-coach-review --prompt "Analyze this Legacy Storm deck. Focus on combo consistency and mana base optimization."

Integration

Genkit Configuration

Flows are configured in /src/ai/genkit.ts:

• Model: gemini-1.5-flash-latest
• Plugin: Google AI

Server Actions

Flows are invoked via server actions in /src/app/actions.ts with 'use server' directive.

Troubleshooting

Flow returns empty output

• Check input format matches expected schema
• Verify API key is set in .env
• Increase model temperature for more creative outputs

Schema validation fails

• Review Zod schema in flow file
• Check for missing required fields in output
• Update flow prompt to be more explicit about output format

Rate limiting

• Add retry logic with exponential backoff
• Cache common test inputs
• Use --delay flag between rapid tests