System Info

Environment Information

Python: 3.11
OS: macOS
llama-prompt-ops: 0.0.7
dspy: Latest compatible version
Installation Method: pip

To capture full environment info, run:

python -m torch.utils.collect_env

Information

• The official example scripts
• My own modified scripts

🐛 Describe the bug

🐛 Describe the bug

The llama-prompt-ops library (version 0.0.7) consistently fails with OptimizationError: 'str' object has no attribute 'kwargs' when attempting to run prompt optimization. This is a persistent library bug that prevents the core optimization functionality from working.

Root Cause: The error occurs in llama_prompt_ops/core/prompt_strategies.py around line 383, where a scope issue causes an optimizer variable to be accessed as a string before it's properly defined as an object with the expected kwargs attribute.

Impact: This bug makes the library completely unusable for its primary purpose - prompt optimization.

# Minimal reproduction command
python run_llama_prompt_ops.py

Results: The optimization process crashes immediately with the kwargs error, preventing any optimization from completing.

Minimal Reproduction Code:

from llama_prompt_ops.core.migrator import PromptMigrator
from llama_prompt_ops.core.prompt_strategies import BasicOptimizationStrategy

# Configuration
config = {
    'model': {
        'name': 'openai/gpt-4o-mini',
        'api_base': 'https://oai.helicone.ai/v1',
        'temperature': 0.0
    },
    'optimization': {
        'strategy': 'basic',
        'max_rounds': 2
    }
}

# This triggers the error
strategy = BasicOptimizationStrategy(
    max_rounds=config['optimization']['max_rounds']
)

migrator = PromptMigrator(strategy=strategy)

# Error occurs here
result = migrator.optimize(
    prompt_data={'text': 'Test prompt', 'inputs': ['input'], 'outputs': ['output']},
    trainset=[],
    save_to_file=False
)

Expected behavior

The optimization should complete successfully and return an optimized prompt with performance metrics, or fail gracefully with a meaningful error message if there are configuration issues.

Instead: The library crashes with an unclear error that points to an internal scope issue, making it impossible to use the core functionality.

Additional Context

• Reproducibility: 100% reproducible across different environments
• Fix Attempts: Multiple comprehensive fixes attempted (scope fixes, patches, clean reinstalls) - all failed
• Severity: High - blocks core functionality
• Workaround: None available - recommend using DSPy directly instead
• Repository: https://github.com/meta-llama/llama-prompt-ops

This appears to be a fundamental library bug requiring maintainer intervention to fix the scope issue in prompt_strategies.py.

Error logs

OptimizationError: 'str' object has no attribute 'kwargs'

The error originates from:
File: /Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages/llama_prompt_ops/core/prompt_strategies.py
Location: Around line 383
Issue: Nested function accessing optimizer.proposer_kwargs.get('tip') when optimizer is still a string

Expected behavior

The optimization should complete successfully and return an optimized prompt with performance metrics, or fail gracefully with a meaningful error message if there are configuration issues.

Instead: The library crashes with an unclear error that points to an internal scope issue, making it impossible to use the core functionality.