matheel 0.3.3

Matheel: A CLI and Python package for source-code similarity detection.

Matheel

Matheel is a function-based Python package and CLI for source-code similarity. It combines semantic embeddings, lexical similarity, chunking, preprocessing, and code-aware metrics in one workflow.

Demos

• Hugging Face Space demo: buelfhood/matheel-framework
• Gradio Colab notebook: Open in Colab
• Examples Colab notebook: Open in Colab

Installation

Use Python 3.10 to 3.12.

Base install:

pip install matheel

Optional extras:

pip install "matheel[chunking]"
pip install "matheel[chunking_code]"
pip install "matheel[metrics]"
pip install "matheel[model2vec]"
pip install "matheel[pylate]"
pip install "matheel[gradio]"
pip install "matheel[all]"
pip install "matheel[dev]"

matheel[all] installs the currently supported optional backends in one command: Chonkie code chunking, metrics runtime dependencies (RUBY graph/tree, TSED, CodeBERTScore), model2vec, PyLate, and the Gradio app dependencies.

Quick Start

The repository root includes sample_pairs.zip, a small Java archive with:

• code_1.java
• code_3_plag.java
• additional plagiarized and non-plagiarized comparisons

CLI archive comparison:

matheel compare sample_pairs.zip \
  --model huggingface/CodeBERTa-small-v1 \
  --feature-weight semantic=0.7 \
  --feature-weight levenshtein=0.3 \
  --threshold 0.2 \
  --num 10

Python pairwise scoring with the sample pair:

from zipfile import ZipFile

from matheel.similarity import calculate_similarity

with ZipFile("sample_pairs.zip") as archive:
    code_a = archive.read("code_1.java").decode("utf-8")
    code_b = archive.read("code_3_plag.java").decode("utf-8")

score = calculate_similarity(
    code_a,
    code_b,
    model_name="huggingface/CodeBERTa-small-v1",
    feature_weights={
        "semantic": 0.7,
        "levenshtein": 0.3,
    },
)

print(round(score, 4))

Supported Scope

Supported languages:

• Chunking and preprocessing are text-first and can run on any source text.
• Code-aware metrics are currently scoped to Java, Python, C, and C++.

Similarity features:

• semantic
• levenshtein
• jaro_winkler
• winnowing
• gst
• code_metric

Code metrics:

• codebleu
• codebleu_ngram
• codebleu_weighted_ngram
• codebleu_syntax
• codebleu_dataflow
• crystalbleu
• ruby
• tsed
• codebertscore

Chunking methods:

• none
• code
• chonkie_token
• chonkie_sentence
• chonkie_recursive
• chonkie_fast

Vector backends:

• auto
• sentence_transformers
• model2vec
• pylate

Single-vector similarity functions:

• cosine
• dot
• euclidean
• manhattan

Sentence Transformers pooling methods:

• mean
• max
• cls
• lasttoken
• mean_sqrt_len_tokens
• weightedmean

auto inspects Hugging Face model metadata and routes to the correct backend when the model exposes a known library.

CLI

Compare a directory or ZIP archive:

matheel compare codes/ \
  --model huggingface/CodeBERTa-small-v1 \
  --vector-backend auto \
  --max-token-length 256 \
  --feature-weight semantic=0.6 \
  --feature-weight levenshtein=0.2 \
  --feature-weight jaro_winkler=0.1 \
  --feature-weight code_metric=0.1 \
  --similarity-function dot \
  --pooling-method max \
  --preprocess-mode basic \
  --chunking-method code \
  --chunk-language python \
  --chunker-option include_line_numbers=true \
  --code-metric codebleu \
  --code-language python \
  --threshold 0.5 \
  --num 25

Run multiple configurations:

matheel compare-suite codes/ runs.json \
  --summary-out results/summary.csv \
  --details-dir results/runs

Python API

Pairwise scoring:

from matheel.similarity import calculate_similarity

code_a = "def add(a, b):\n    return a + b\n"
code_b = "def add(x, y):\n    return x + y\n"

score = calculate_similarity(
    code_a,
    code_b,
    model_name="huggingface/CodeBERTa-small-v1",
    vector_backend="auto",
    max_token_length=256,
    similarity_function="dot",
    pooling_method="max",
    preprocess_mode="basic",
    chunking_method="code",
    chunk_language="python",
    code_metric="codebleu",
    code_language="python",
    feature_weights={"semantic": 0.5, "code_metric": 0.5},
)

Directory or ZIP ranking:

from matheel.similarity import get_sim_list

results = get_sim_list(
    "sample_pairs.zip",
    model_name="huggingface/CodeBERTa-small-v1",
    threshold=0.4,
    number_results=10,
    vector_backend="auto",
    max_token_length=256,
    chunking_method="chonkie_token",
    chunk_size=120,
    chunk_overlap=20,
    similarity_function="cosine",
    pooling_method="mean",
    feature_weights={
        "semantic": 0.7,
        "levenshtein": 0.15,
        "jaro_winkler": 0.05,
        "winnowing": 0.05,
        "gst": 0.05,
    },
)

print(results.head())

Docs

• Docs folder landing page: docs/README.md
• Canonical docs index: docs/index.md
• Quick usage: docs/usage.md
• Preprocessing: docs/preprocessing.md
• Chunking: docs/chunking.md
• Vectors and routing: docs/vectors.md
• Lexical metrics and baselines: docs/lexical.md
• Code metrics: docs/code_metrics.md
• Comparison suite: docs/comparison_suite.md

Examples

• Colab walkthrough: examples/matheel_examples_colab.ipynb
• Quick archive check: examples/sample_pairs_demo.py
• Preprocessing: examples/preprocessing_demo.py
• Chunking: examples/chunking_demo.py
• Vector backends: examples/vectors_demo.py
• Lexical metrics and baselines: examples/lexical_demo.py
• Code metrics: examples/code_metrics_demo.py
• Comparison suite: examples/comparison_suite_demo.py

All examples use the same sample pair from sample_pairs.zip: code_1.java and code_3_plag.java.

Gradio

The Gradio demo stays in gradio_app/ and is aligned with the Hugging Face Space setup. The UI supports embeddings, lexical metrics, baseline algorithms (Winnowing and GST), and the code-aware metrics (CodeBLEU, CrystalBLEU, RUBY, TSED, CodeBERTScore), with metric-specific advanced fields. The core package and CLI can read either a ZIP archive or a directory; the Gradio upload flow remains ZIP-based.

Acknowledgments

Matheel builds on several open-source libraries:

• Sentence Transformers
• Chonkie
• model2vec
• PyLate
• RapidFuzz
• tree-sitter-language-pack
• NetworkX
• APTED
• bert-score
• Gradio

The project also depends on the standard scientific Python stack and related tooling, including NumPy, pandas, Click, SentencePiece, and func-timeout.

License

This project is licensed under the MIT License.