TanML: Automated Model Validation Toolkit for Tabular Machine Learning Models

TanML bridges the gap between data science tools (for building models) and governance requirements (for validating them). It's not just an ML library—it's a validation workflow with built-in documentation.

---

🚀 See TanML in Action (2-Minute Walkthrough)

---

📖 Documentation

For detailed guides, API reference, and tutorials, visit our official documentation:

👉 https://tdlabs-ai.github.io/tanml/

---

Table of Contents

• Why TanML?
• Install
• Remote Access (SSH)
• Quick Start (UI)
• Library Usage (Python)
• Testing
• Optional CLI Flags
• Reports
• Data Privacy
• Troubleshooting
• Contributing
• License & Citation

---

Why TanML?

• End-to-end workflow: Data Profiling → Data Preprocessing → Feature Power Ranking → Model Development → Evaluation → Reports—all in one UI.
• Audit-ready reports: Generate editable .docx documents for stakeholders and compliance reviews.
• Built for regulated industries: Designed for MRM, credit risk, insurance, and SR 11-7 contexts.
• No code required: Fully UI-driven—no Python knowledge needed.
• Robust evaluation: Drift detection, stress testing, SHAP explainability, cluster coverage.
• Works with your stack: Scikit-learn, XGBoost, LightGBM, CatBoost, Statsmodels.

Comparison to Similar Tools

Feature	TanML	Evidently AI	Deepchecks	AutoML (PyCaret/AutoGluon)	YData Profiling
Primary Focus	Model Risk Management	Monitoring	Testing / CI	Model Training	Data EDA Only
Scope	Data + Model + Governance	Data + Model	Data + Model	Model Building	Data Only
Output	Audit-Ready .docx	Dashboards	Reports	Models	HTML Report
Drift Logic	PSI & KS (Regulatory)	Stat. Tests	Multiple	N/A	Warnings

---

Install

We strongly recommend using a virtual environment to isolate dependencies:

# Create and activate virtual environment
python -m venv .venv
source .venv/bin/activate  # Linux/Mac
# .venv\Scripts\activate   # Windows

# Install TanML
pip install tanml

# Upgrade an existing installation
python -m pip install --upgrade tanml

Run TanML Locally

If you are running TanML directly on your own machine, start the UI with:

tanml ui

If the tanml command is not available in your shell, you can also try:

python -m tanml ui

Then open the following in your browser: http://localhost:8501

Remote Access (SSH)

If you are using standard SSH from a terminal, you may need to forward port 8501 manually. If you are using VS Code Remote SSH, VS Code may detect and forward the TanML UI port automatically.

Standard SSH from a terminal

From your local machine, connect to the remote server with port forwarding enabled:

ssh -L 8501:localhost:8501 user@remote-server-ip

Then, on the remote server, run:

tanml ui

After that, open the following in your local browser:

http://localhost:8501

VS Code Remote SSH

If you are connected through VS Code Remote SSH, run tanml ui in the remote terminal. VS Code may automatically detect port 8501 and offer to forward it. If it does not, open the Ports panel in VS Code, choose Forward a Port, enter 8501, and then open the forwarded local address in your browser.

---

Quick Start (UI)

tanml ui

• Opens at http://127.0.0.1:8501
• Upload limit ~2 GB (preconfigured)
• Telemetry disabled by default

Library Usage (Python)

You can also use TanML's analysis engine directly in your Python scripts:

import pandas as pd
from tanml.analysis import analyze_drift

# Load your data
train_df = pd.read_csv("train.csv")
test_df = pd.read_csv("test.csv")

# Run drift analysis (PSI, KS Test, etc.)
results = analyze_drift(train_df, test_df)
print(results)

---

Testing

TanML uses pytest for automated unit testing. To run tests locally:

# 1. Navigate to the project root
cd tanml

# 2. Install with dev dependencies
pip install -e ".[dev]"

# 3. Run all tests
pytest tests/

We also use ruff for linting and mypy for type checking. These are integrated into our CI pipeline.

---

Optional CLI Flags

Most users just run tanml ui. These help on teams/servers:

# Share on LAN
tanml ui --public

> [!WARNING]
> Using `--public` makes the UI accessible to anyone on your network. Only use this on trusted networks (like a VPN or secure office LAN) and never on public Wi-Fi.

# Different port
tanml ui --port 9000

# Headless (server/CI; no auto-open browser)
tanml ui --headless

# Larger limit (e.g., 2 GB)
tanml ui --max-mb 2048

Env var equivalents (Linux/macOS bash):

TANML_SERVER_ADDRESS=0.0.0.0 TANML_PORT=9000 TANML_MAX_MB=2048 tanml ui

Windows PowerShell:

$env:TANML_SERVER_ADDRESS="0.0.0.0"; $env:TANML_PORT="9000"; $env:TANML_MAX_MB="2048"; tanml ui

Defaults: address 127.0.0.1, port 8501, limit 2048 MB, telemetry OFF.

---

Reports

TanML generates audit ready reports (.docx) programmatically:

• Model Development Report — Cross-validation metrics, diagnostics, and performance summary
• Model Evaluation Report — Train/Test comparison, drift analysis, stress testing, SHAP explainability
• Feature Power Ranking Report — Feature importance scores, correlation analysis

Reports are generated via tanml/ui/reports/generators.py and exported directly from the UI.

---

Data Privacy

• TanML runs locally; no data is sent to external services.
• Telemetry is disabled by default (and can be forced off via --no-telemetry).
• Security Note: Running with the --public flag exposes the app to your local network. Ensure you are on a secure connection.
• UI artifacts and reports are written under ./tanml_runs/<session>/ in your working directory.

---

Troubleshooting

• Page didn’t open? Visit http://127.0.0.1:8501 or run tanml ui --port 9000.
• Large CSVs are slow/heavy? Prefer Parquet; CSV → DataFrame can use several GB RAM.
• Artifacts missing? Check ./tanml_runs/<session>/artifacts/.
• Corporate networks: use tanml ui --public to share on LAN.

---

Contributing

We welcome issues and PRs! See CONTRIBUTING.md for detailed setup instructions.

1. Clone & Setup: git clone https://github.com/tdlabs-ai/tanml.git && cd tanml
2. Environment: python -m venv .venv && source .venv/bin/activate
3. Install: pip install -e .[dev]
4. Test: pytest

---

License & Citation

License: MIT. See LICENSE.
SPDX-License-Identifier: MIT

© 2026 Tanmay Sah and Dolly Sah. You may use, modify, and distribute this software with appropriate attribution.

How to cite

If TanML helps your work or publications, please cite:

Sah, T., & Sah, D. (2026). TanML: Automated Model Validation Toolkit for Tabular Machine Learning Models [Software]. Zenodo. https://doi.org/10.5281/zenodo.17317165

Or in BibTeX (version-agnostic):

@software{tanml_2026,
  author       = {Sah, Tanmay and Sah, Dolly},
  title        = {TanML: Automated Model Validation Toolkit for Tabular Machine Learning Models},
  year         = {2026},
  publisher    = {Zenodo},
  doi          = {10.5281/zenodo.17317165},
  url          = {https://doi.org/10.5281/zenodo.17317165},
  license      = {MIT}
}

A machine-readable citation file (CITATION.cff) is included for citation tools and GitHub’s “Cite this repository” button.

---

AI Disclosure

Portions of this codebase (including tests, documentation, and the data_loader module) were refactored and generated with the assistance of Large Language Models (LLMs). All AI-generated contributions have been reviewed and verified by the human maintainers.