Skip to content

~95% on SimpleQA (e.g. Qwen3.6-27B on a 3090). Supports all local and cloud LLMs (llama.cpp, Ollama, Google, ...). 10+ search engines - arXiv, PubMed, your private documents. Everything Local & Encrypted.

6.5 KB
Raw Blame
#!/usr/bin/env python
"""
Optimization Example with Gemini 2.0 Flash via OpenRouter.

This script demonstrates how to run parameter optimization using the Gemini 2.0 Flash
model via OpenRouter.

Usage:
    # Install dependencies with PDM
    cd /path/to/local-deep-research
    pdm install

    # Set your OpenRouter API key
    export OPENAI_ENDPOINT_API_KEY="your_openrouter_api_key"

    # Run the script with PDM
    pdm run python examples/optimization/gemini_optimization.py
"""

import argparse
import json
import os
import sys
from datetime import datetime, timezone
from pathlib import Path

from loguru import logger

# Import the optimization functionality
from local_deep_research.benchmarks.optimization import (
    optimize_for_quality,
    optimize_for_speed,
    optimize_parameters,
)


def setup_gemini_config(api_key=None):
    """
    Create a configuration for using Gemini via OpenRouter.

    Args:
        api_key: OpenRouter API key. If None, will try to get from environment.

    Returns:
        Dictionary with Gemini configuration.
    """
    # Get API key from argument or environment
    if not api_key:
        api_key = os.environ.get("OPENAI_ENDPOINT_API_KEY")
        if not api_key:
            api_key = os.environ.get("LDR_LLM__OPENAI_ENDPOINT_API_KEY")

    if not api_key:
        logger.error("No API key found. Please provide an OpenRouter API key.")
        return None

    return {
        "model_name": "google/gemini-2.0-flash-001",  # OpenRouter format for Gemini
        "provider": "openai_endpoint",  # Use OpenRouter as endpoint
        "openai_endpoint_url": "https://openrouter.ai/api/v1",
        "api_key": api_key,
    }


def main():
    # Parse arguments
    parser = argparse.ArgumentParser(
        description="Run optimization with Gemini 2.0 Flash via OpenRouter"
    )
    parser.add_argument(
        "--api-key",
        help="OpenRouter API key. If not provided, will try to use from environment.",
    )
    parser.add_argument(
        "--mode",
        choices=["balanced", "speed", "quality"],
        default="balanced",
        help="Optimization mode (default: balanced)",
    )
    parser.add_argument(
        "--trials",
        type=int,
        default=3,
        help="Number of optimization trials (default: 3)",
    )
    parser.add_argument(
        "--output-dir",
        default=None,
        help="Directory to save results (default: auto-generated)",
    )
    args = parser.parse_args()

    # Set up Gemini configuration
    gemini_config = setup_gemini_config(args.api_key)
    if not gemini_config:
        return 1

    # Create timestamp for unique output directory
    timestamp = datetime.now(timezone.utc).strftime("%Y%m%d_%H%M%S")
    if args.output_dir:
        output_dir = args.output_dir
    else:
        output_dir = str(
            Path("examples")
            / "optimization"
            / "results"
            / f"gemini_opt_{timestamp}"
        )
    Path(output_dir).mkdir(parents=True, exist_ok=True)

    print(
        f"Starting optimization with Gemini 2.0 Flash - results will be saved to {output_dir}"
    )
    print(
        f"Using model: {gemini_config['model_name']} via {gemini_config['provider']}"
    )

    # Set environment variables to ensure proper API access
    os.environ["OPENAI_ENDPOINT_API_KEY"] = gemini_config["api_key"]
    os.environ["LDR_LLM__OPENAI_ENDPOINT_API_KEY"] = gemini_config["api_key"]
    os.environ["OPENAI_ENDPOINT_URL"] = gemini_config["openai_endpoint_url"]
    os.environ["LDR_LLM__OPENAI_ENDPOINT_URL"] = gemini_config[
        "openai_endpoint_url"
    ]
    os.environ["LDR_LLM__PROVIDER"] = gemini_config["provider"]
    os.environ["LDR_LLM__MODEL"] = gemini_config["model_name"]

    # Create a very simple parameter space for quick demonstration
    param_space = {
        "iterations": {
            "type": "int",
            "low": 1,
            "high": 2,
            "step": 1,
        },
        "questions_per_iteration": {
            "type": "int",
            "low": 1,
            "high": 2,
            "step": 1,
        },
        "search_strategy": {
            "type": "categorical",
            "choices": ["rapid", "source_based"],  # Limited choices for speed
        },
    }

    # Run optimization based on selected mode
    query = "Recent developments in fusion energy research"

    try:
        if args.mode == "speed":
            print("\n=== Running speed-focused optimization with Gemini ===")
            best_params, best_score = optimize_for_speed(
                query=query,
                param_space=param_space,
                n_trials=args.trials,
                model_name=gemini_config["model_name"],
                provider=gemini_config["provider"],
                output_dir=output_dir,
            )
        elif args.mode == "quality":
            print("\n=== Running quality-focused optimization with Gemini ===")
            best_params, best_score = optimize_for_quality(
                query=query,
                param_space=param_space,
                n_trials=args.trials,
                model_name=gemini_config["model_name"],
                provider=gemini_config["provider"],
                output_dir=output_dir,
            )
        else:  # balanced
            print("\n=== Running balanced optimization with Gemini ===")
            best_params, best_score = optimize_parameters(
                query=query,
                param_space=param_space,
                n_trials=args.trials,
                model_name=gemini_config["model_name"],
                provider=gemini_config["provider"],
                output_dir=output_dir,
                metric_weights={"quality": 0.5, "speed": 0.5},
            )

        print(f"Best parameters: {best_params}")
        print(f"Best score: {best_score:.4f}")

        # Save summary to JSON
        summary = {
            "timestamp": timestamp,
            "mode": args.mode,
            "model": gemini_config["model_name"],
            "provider": gemini_config["provider"],
            "best_parameters": best_params,
            "best_score": float(best_score),
        }

        with open(
            Path(output_dir) / "gemini_optimization_summary.json",
            "w",
            encoding="utf-8",
        ) as f:
            json.dump(summary, f, indent=2)

        print(f"\nOptimization complete! Results saved to {output_dir}")
        print(f"Recommended parameters for {args.mode} mode: {best_params}")

    except Exception:
        logger.exception("Error during optimization")
        return 1

    return 0


if __name__ == "__main__":
    sys.exit(main())