A feedback loop workflow where LLM-generated outputs are evaluated, refined, and optimized iteratively to improve accuracy and relevance.

​
Quick Start

1

Install Package

First, install the PraisonAI Agents package:
pip install praisonaiagents
2

Set API Key

Set your OpenAI API key as an environment variable in your terminal:
export OPENAI_API_KEY=your_api_key_here
3

Create a file

Create a new file app.py with the basic setup:
from praisonaiagents import Agent, Task, PraisonAIAgents

# Create generator and evaluator agents
generator = Agent(
    name="Generator",
    role="Solution generator",
    goal="Generate initial solutions and incorporate feedback",
    instructions=(
        "1. Look at the context from previous tasks.\n"
        "2. If you see that you have already produced 2 points, then add another 2 new points "
        "   so that the total becomes 10.\n"
        "3. Otherwise, just produce the first 2 points.\n"
        "4. Return only the final list of points, with no extra explanation."
    )
)

evaluator = Agent(
    name="Evaluator",
    role="Solution evaluator",
    goal="Evaluate solutions and provide improvement feedback",
    instructions=(
        "1. Count how many lines in the response start with a number and a period (like '1. ' or '2. ').\n"
        "2. If there are 10 or more, respond with 'done'.\n"
        "3. Otherwise, respond with 'more'.\n"
        "4. Return only the single word: 'done' or 'more'."
    )
)

# Create tasks for the feedback loop
generate_task = Task(
    name="generate",
    description="Write 2 points about AI incuding if anything exiting from previous points",
    expected_output="2 points",
    agent=generator,
    is_start=True,
    task_type="decision",
    next_tasks=["evaluate"]
)

evaluate_task = Task(
    name="evaluate",
    description="Check if there are 10 points about AI",
    expected_output="more or done",
    agent=evaluator,
    next_tasks=["generate"],
    context=[generate_task],
    task_type="decision",
    condition={
        "more": ["generate"],  # Continue to generate
        "done": [""]  # Exit when optimization complete
    }
)

# Create workflow manager
workflow = PraisonAIAgents(
    agents=[generator, evaluator],
    tasks=[generate_task, evaluate_task],
    process="workflow",
    verbose=True
)

# Run optimization workflow
results = workflow.start()

# Print results
print("\nEvaluator-Optimizer Results:")
for task_id, result in results["task_results"].items():
    if result:
        print(f"Task {task_id}: {result.raw}")
4

Start Agents

Type this in your terminal to run your agents:
python app.py
Requirements
  • Python 3.10 or higher
  • OpenAI API key. Generate OpenAI API key here. Use Other models using this guide.
  • Basic understanding of Python

​
Understanding Evaluator-Optimizer

What is Evaluator-Optimizer?

Evaluator-Optimizer pattern enables:
  • Iterative solution generation and refinement
  • Automated quality evaluation
  • Feedback-driven optimization
  • Continuous improvement loops

​
Features

Solution Generation

Generate solutions based on requirements and feedback.

Quality Evaluation

Automatically assess solution quality and completeness.

Feedback Loop

Implement iterative improvement through feedback cycles.

Process Control

Monitor and control the optimization process.

​
Configuration Options

# Create a generator agent
generator = Agent(
    name="Generator",
    role="Solution generator",
    goal="Generate and improve solutions",
    instructions="Step-by-step instructions for generation",
    verbose=True  # Enable detailed logging
)

# Create an evaluator agent
evaluator = Agent(
    name="Evaluator",
    role="Solution evaluator",
    goal="Evaluate and provide feedback",
    instructions="Evaluation criteria and feedback format"
)

# Create tasks with feedback loop
generate_task = Task(
    name="generate",
    description="Generate solution",
    agent=generator,
    is_start=True,
    task_type="decision",
    next_tasks=["evaluate"]
)

evaluate_task = Task(
    name="evaluate",
    description="Evaluate solution",
    agent=evaluator,
    context=[generate_task],
    task_type="decision",
    condition={
        "more": ["generate"],  # Continue optimization
        "done": [""]  # Exit when complete
    }
)

​
Troubleshooting

Generation Issues

If generation is not improving:
  • Review generator instructions
  • Check feedback integration
  • Enable verbose mode for debugging

Evaluation Flow

If evaluation cycle is incorrect:
  • Verify evaluation criteria
  • Check condition mappings
  • Review feedback loop connections

​
Next Steps

AutoAgents

Learn about automatically created and managed AI agents

Mini Agents

Explore lightweight, focused AI agents
For optimal results, ensure your generator instructions and evaluation criteria are clear and well-defined to achieve the desired optimization outcomes.