Plexe

<div align="center">

plexe ✨

<img src="https://raw.githubusercontent.com/plexe-ai/plexe/main/resources/backed-by-yc.png" alt="backed-by-yc" width="20%">

Build machine learning models using natural language.

Quickstart | Features | Installation | Documentation

<br>

plexe lets you create machine learning models by describing them in plain language. Simply explain what you want, provide a dataset, and the AI-powered system builds a fully functional model through an automated agentic approach. Also available as a managed cloud service.

<br>

Watch the demo on YouTube:

</div>

1. Quickstart

Installation

pip install plexe
export OPENAI_API_KEY=<your-key>
export ANTHROPIC_API_KEY=<your-key>

Using plexe

Provide a tabular dataset (Parquet, CSV, ORC, or Avro) and a natural language intent:

python -m plexe.main \
    --train-dataset-uri data.parquet \
    --intent "predict whether a passenger was transported" \
    --max-iterations 5

from plexe.main import main
from pathlib import Path

best_solution, metrics, report = main(
    intent="predict whether a passenger was transported",
    data_refs=["train.parquet"],
    max_iterations=5,
    work_dir=Path("./workdir"),
)
print(f"Performance: {best_solution.performance:.4f}")

2. Features

2.1. 🤖 Multi-Agent Architecture

The system uses 14 specialized AI agents across a 6-phase workflow to:

• Analyze your data and identify the ML task
• Select the right evaluation metric
• Search for the best model through hypothesis-driven iteration
• Evaluate model performance and robustness
• Package the model for deployment

2.2. 🎯 Automated Model Building

Build complete models with a single call. Plexe supports XGBoost, CatBoost, LightGBM, Keras, and PyTorch for tabular data:

best_solution, metrics, report = main(
    intent="predict house prices based on property features",
    data_refs=["housing.parquet"],
    max_iterations=10,                    # Search iterations
    allowed_model_types=["xgboost"],      # Or let plexe choose
    enable_final_evaluation=True,         # Evaluate on held-out test set
)

Run python -m plexe.main --help for all CLI options.

The output is a self-contained model package at work_dir/model/ (also archived as model.tar.gz). The package has no dependency on plexe — build the model with plexe, deploy it anywhere:

model/
├── artifacts/          # Trained model + feature pipeline (pickle)
├── src/                # Inference predictor, pipeline code, training template
├── schemas/            # Input/output JSON schemas
├── config/             # Hyperparameters
├── evaluation/         # Metrics and detailed analysis reports
├── model.yaml          # Model metadata
└── README.md           # Usage instructions with example code

2.3. 🐳 Batteries-Included Docker Images

Run plexe with everything pre-configured — PySpark, Java, and all dependencies included. A Makefile is provided for common workflows:

make build          # Build the Docker image
make test-quick     # Fast sanity check (~1 iteration)
make run-titanic    # Run on Spaceship Titanic dataset

Or run directly:

docker run --rm \
    -e OPENAI_API_KEY=$OPENAI_API_KEY \
    -e ANTHROPIC_API_KEY=$ANTHROPIC_API_KEY \
    -v $(pwd)/data:/data -v $(pwd)/workdir:/workdir \
    plexe:py3.12 python -m plexe.main \
        --train-dataset-uri /data/dataset.parquet \
        --intent "predict customer churn" \
        --work-dir /workdir \
        --spark-mode local

A config.yaml in the project root is automatically mounted. A Databricks Connect image is also available: docker build --target databricks .

2.4. ⚙️ YAML Configuration

Customize LLM routing, search parameters, Spark settings, and more via a config file:

# config.yaml
max_search_iterations: 5
allowed_model_types: [xgboost, catboost]
spark_driver_memory: "4g"
hypothesiser_llm: "openai/gpt-5-mini"
feature_processor_llm: "anthropic/claude-sonnet-4-5-20250929"

CONFIG_FILE=config.yaml python -m plexe.main ...