Skip to content

JGITSol/universal_solver

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

18 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Universal Solver

Test Coverage License: MIT


Universal Solver is a modular, extensible platform for advanced mathematical problem solving, symbolic regression, and AI-driven research workflows. It integrates state-of-the-art models, ensemble methods, and collaborative tools to accelerate research and innovation in mathematics, science, and engineering.

Key Features

  • Advanced Math Ensemble Solver: Combines multiple state-of-the-art tools and models—including LangChain, Ollama, OpenRouter, Google Gemini, and SymPy—for symbolic mathematics and regression.
  • Industry-Standard Math Benchmarking: Supports a wide range of math benchmarks (MATH, GSM8K, MathQA, ASDiv, SVAMP, AQUA-RAT, MiniF2F, and more) via HuggingFace Datasets.
  • Flexible Interfaces: Provides both a Command-Line Interface (CLI) and a modern graphical user interface (GUI) built with CustomTkinter, as well as Jupyter/Colab notebook support for collaborative and cloud-based workflows.
  • Extensible Architecture: Easily add new models, solvers, and research workflows with a plugin-friendly architecture.
  • Comprehensive Benchmarking and Reporting: Run large-scale benchmarks, export results to Excel/Parquet, and upload to cloud storage (GCP, Azure, Kaggle).
  • Modern Python Tooling: Fully type-checked, linted, and covered by automated tests. Includes development tools for formatting, linting, and static analysis.

Project Structure

adv_resolver_math/         # Advanced math ensemble solver (LangChain, Ollama, OpenRouter, Gemini, SymPy, etc.)
KAN/                      # Symbolic regression with Kolmogorov-Arnold Networks (KAN)
benchmark_datasets.py     # Loader for standard math benchmarks
benchmark_cli.py          # CLI for running solver benchmarks
benchmark_showcase_colab.ipynb # Colab/cloud notebook for benchmarking and sharing
collab_training_ntbks/    # Collaborative model training notebooks
docs/                     # Documentation, guides, and testing
model/                    # Model state, configs, and history
project_guidelines/       # Hackathon and project guidelines, agent specs
math_cache/               # Exported math data and cache
tests/                    # Test suite (pytest compatible)
universal_solver_gui.py   # Modern GUI for solver interaction
... (see [Project Overview](docs/UNIVERSAL_SOLVER_PROJECT_OVERVIEW.md))

Installation Prerequisites Python 3.8+ pip Clone and Prepare sh CopyInsert git clone universal_solver cd universal_solver Setup Environments For Advanced Math Ensemble sh CopyInsert cd adv_resolver_math python -m venv adv_res_venv adv_res_venv\Scripts\activate # On Windows pip install -r requirements.txt For KAN Module sh CopyInsert cd ../KAN python -m venv venv venv\Scripts\activate pip install -r requirements.txt Install Core Dependencies Alternatively, install all core dependencies in the project root:

sh CopyInsert pip install -r requirements.txt Or for development:

sh CopyInsert pip install -e .[dev] Configuration Create a .env file in the project root to store API keys and configuration parameters for external services (e.g., OpenAI, Gemini, etc.).

Usage CLI Run benchmarks or solve problems via the command line:

sh CopyInsert python benchmark_cli.py --help GUI Launch the graphical interface:

sh CopyInsert python universal_solver_gui.py Jupyter/Colab Use collaborative notebooks in collab_training_ntbks/ or benchmark_showcase_colab.ipynb for cloud-based workflows.

Example Workflow Select a math problem or benchmark dataset. Choose solver options (ensemble, symbolic, neural, etc.). Process the problem and review intermediate logs and results. Use voting and debugging panels (GUI) for transparency and inspection. Export results and reports as needed. Testing & Quality Assurance Run all tests: sh CopyInsert pytest --cov=adv_resolver_math --cov-report=html Code is formatted with black, linted with flake8, and type-checked with mypy. Dependencies Core dependencies include:

numpy, sympy, torch, scikit-learn, sentence-transformers, transformers, pandas, matplotlib, seaborn, plotly, langchain, pykan, customtkinter, rich, requests, and more. Full list in requirements.txt and setup.py.

License MIT License

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published