FluxLoop CLI
Command-line interface for managing FluxLoop projects, generating inputs, and running agent simulations.
Installation
pip install fluxloop-cli
Version
Current version: 0.2.30
Features
- 🎯 Project Management: Initialize and configure simulation projects
- 📝 Input Generation: Create test input variations using LLM or deterministic strategies
- 🧪 Experiment Execution: Run batch simulations with configurable iterations
- 💬 Multi-Turn Conversations: Automatically extend experiments into dynamic dialogues with AI supervisor
- 📊 Result Parsing & Evaluation: Convert raw artifacts to human-readable formats and score with evaluators
- 🔴 Recording Mode: Capture and replay complex function arguments
- ⚙️ Configuration: Structured YAML-based configuration system
- 🩺 Environment Diagnostics: Verify installation and detect configuration issues
Quick Example
# Initialize a new project
fluxloop init project --name my-agent
cd fluxloop/my-agent
# Generate 50 input variations
fluxloop generate inputs --limit 50
# Run experiment
fluxloop run experiment
# Parse results
fluxloop parse experiment experiments/exp_*/
# Evaluate results (writes evaluation_report/report.html by default)
fluxloop evaluate experiment experiments/exp_*/
# Check environment
fluxloop doctor
Key Commands
Project Initialization
fluxloop init project --name <project-name>
Creates a new FluxLoop project with:
configs/directory (project, input, simulation, evaluation configs).envfile for environment variablesexamples/with sample agent codeinputs/,recordings/,experiments/directories
Input Generation
fluxloop generate inputs [--limit 100] [--mode llm]
Generate input variations from base inputs defined in configs/input.yaml.
Experiment Execution
fluxloop run experiment [--iterations 10]
# Multi-turn conversations (LLM-driven)
fluxloop run experiment --multi-turn --max-turns 12 --auto-approve
# Multi-turn scripted playback (deterministic)
# Set supervisor.provider to 'mock' and supply scripted_questions in simulation.yaml
fluxloop run experiment --multi-turn --supervisor-provider mock
Run simulation experiment using configuration from configs/simulation.yaml. Enable multi-turn mode to automatically extend single inputs into dynamic conversations. Use provider: openai for AI-driven follow-ups or provider: mock for scripted question playback.
Recording Mode
fluxloop record enable # Enable argument recording
fluxloop record disable # Disable recording
fluxloop record status # Check recording state
Configuration
fluxloop config set-llm <provider> <api-key> [--model <model>]
Update LLM provider settings in configs/input.yaml.
Configuration Files (v0.2.0)
FluxLoop CLI uses a multi-file configuration structure:
| File | Purpose |
|---|---|
configs/project.yaml | Project metadata, service context, collector settings |
configs/input.yaml | Personas, base inputs, LLM settings |
configs/simulation.yaml | Runner, iterations, multi-turn settings |
configs/evaluation.yaml | Evaluator definitions |
What's Next?
- Installation - Detailed installation guide
- Project Setup - Initialize your first project
- Commands Reference - Full command documentation
- Doctor Command - Diagnose environment and configuration
- Configuration - Config file reference
- Multi-Turn Workflow - Run dynamic conversations
- Runner Targets - Connect your code to simulations
Need help? Run fluxloop --help or check GitHub Issues.