generate Command

Generate input variations from base inputs using LLM or deterministic strategies.

Usage

fluxloop generate inputs [OPTIONS]

Options

Option	Description	Default
--limit	Number of inputs to generate	50
--mode	Generation mode: llm or deterministic	From config
--output	Output file path	inputs/generated.yaml
--overwrite	Overwrite existing file	false

Examples

Basic Generation

Generate 50 inputs using LLM:

fluxloop generate inputs --limit 50

Deterministic Mode

Generate inputs without LLM:

fluxloop generate inputs --mode deterministic --limit 100

Custom Output

Save to custom location:

fluxloop generate inputs --output my-inputs.yaml --limit 30

Overwrite Existing

Replace existing generated inputs:

fluxloop generate inputs --overwrite --limit 100

How It Works

1. Reads Configuration

Loads personas and base inputs from configs/input.yaml:

personas:
  - name: novice_user
    description: New to the system

base_inputs:
  - input: "How do I start?"
    persona: novice_user

2. Applies Generation Strategies

LLM Mode:

• Rephrase: Different wording, same meaning
• Verbose: Longer, more detailed versions
• Concise: Shorter, direct versions
• Error-prone: Typos, grammar mistakes

Deterministic Mode:

• Pattern-based transformations
• No LLM API calls required

3. Outputs Generated File

Creates inputs/generated.yaml:

generated_inputs:
  - input: "What are the steps to get started?"
    source: "How do I start?"
    strategy: rephrase
    persona: novice_user
  
  - input: "how do i strt?"
    source: "How do I start?"
    strategy: error_prone
    persona: novice_user

Configuration

Set LLM provider in configs/input.yaml:

input_generation:
  mode: llm
  llm:
    provider: openai        # or anthropic
    model: gpt-4o-mini
    temperature: 0.7
  strategies:
    - rephrase
    - verbose
    - concise
    - error_prone

API Key Setup

Before generating with LLM:

# Set API key
fluxloop config set-llm openai sk-your-key --model gpt-4o-mini

# Or add to .env
echo "OPENAI_API_KEY=sk-your-key" >> .env

Best Practices

1. Start small: Generate 10-20 inputs first to test
2. Review outputs: Check quality before generating large batches
3. Mix strategies: Use all strategies for diversity
4. Version control: Commit generated inputs for reproducibility

Troubleshooting

"No API key found"

Set your LLM provider API key:

fluxloop config set-llm openai sk-your-key

"Generation failed"

Check your internet connection and API key validity.

"File already exists"

Use --overwrite flag or delete existing file:

rm inputs/generated.yaml
fluxloop generate inputs

Next Steps

• Run Experiment - Run simulation with generated inputs
• Input Configuration - Configure generation settings