2025.12.21193050

ai-benchmark-analyzer

A Python package that analyzes and summarizes AI benchmark results from unstructured text descriptions. It extracts structured insights like top-performing models, key metrics, and comparative analysis using pattern-matching techniques.

---

📦 Installation

Install via pip:

pip install ai_benchmark_analyzer

---

🚀 Quick Start

Basic Usage (Default LLM7)

from ai_benchmark_analyzer import ai_benchmark_analyzer

user_input = """
Model X achieved 92.3% accuracy on SOTA dataset with 128B parameters.
Model Y scored 89.1% but used only 7B parameters.
"""
response = ai_benchmark_analyzer(user_input)
print(response)

Custom LLM Integration

Replace the default ChatLLM7 with your preferred LLM (e.g., OpenAI, Anthropic, Google):

OpenAI Example

from langchain_openai import ChatOpenAI
from ai_benchmark_analyzer import ai_benchmark_analyzer

llm = ChatOpenAI()
response = ai_benchmark_analyzer(user_input, llm=llm)

Anthropic Example

from langchain_anthropic import ChatAnthropic
from ai_benchmark_analyzer import ai_benchmark_analyzer

llm = ChatAnthropic()
response = ai_benchmark_analyzer(user_input, llm=llm)

Google Generative AI Example

from langchain_google_genai import ChatGoogleGenerativeAI
from ai_benchmark_analyzer import ai_benchmark_analyzer

llm = ChatGoogleGenerativeAI()
response = ai_benchmark_analyzer(user_input, llm=llm)

---

🔧 Parameters

---

🔗 Default LLM: LLM7

The package uses ChatLLM7 by default. Free tier rate limits are sufficient for most use cases. For higher limits:

• Set LLM7_API_KEY environment variable.
• Pass the key directly: ai_benchmark_analyzer(api_key="your_key").

Get a free API key at LLM7 Token.

---

📝 Output Format

The function returns a list of structured strings matching a predefined regex pattern, ensuring consistent and reliable output formatting. Example output:

[
    "Model: ModelX, Metric: Accuracy, Value: 92.3%, Dataset: SOTA, Parameters: 128B",
    "Model: ModelY, Metric: Accuracy, Value: 89.1%, Dataset: SOTA, Parameters: 7B"
]

---

🛠️ Customization

• Pattern Matching: The output adheres to a regex pattern (defined in prompts.py). Modify this file to adjust expected output formats.
• LLM Prompts: System/human prompts are configurable via prompts.py.

---

📜 License

MIT

---

📢 Support & Issues

For bugs/feature requests, open an issue on GitHub.

---

👤 Author

Eugene Evstafev (@chigwell) 📧 hi@euegne.plus