TrustyAI Service

👋 The TrustyAI Service is intended to be a hub for all kinds of Responsible AI workflows, such as explainability, drift, and Large Language Model (LLM) evaluation. Designed as a REST server wrapping a core Python library, the TrustyAI service is intended to be a tool that can operate in a local environment, a Jupyter Notebook, or in Kubernetes.

---

Native Algorithms

📈Drift 📉

• Fourier Maximum Mean Discrepancy (FourierMMD)
• Jensen-Shannon
• Approximate Kolmogorov–Smirnov Test
• Kolmogorov–Smirnov Test (KS-Test)
• Meanshift

⚖️ Fairness ⚖️

• Statistical Parity Difference
• Disparate Impact Ratio
• Average Odds Ratio (WIP)
• Average Predictive Value Difference (WIP)
• Individual Consistency (WIP)

---

Imported Algorithms/Libraries

🔬Explainability 🔬

• LIME (WIP)
• SHAP (WIP)

📋 LLM Evaluation 📋

• LM-Evaluation-Harness

---

📦 Building 📦

Locally

uv pip install ".[$EXTRAS]"

Container

podman build -t $IMAGE_NAME --build-arg EXTRAS="$EXTRAS" .

Available Extras

Pass these extras as a comma separated list, e.g., "mariadb,protobuf"

• protobuf: To process model inference data from ModelMesh models, you can install with protobuf support. Otherwise, only KServe models will be supported.
• eval: To enable the Language Model Evaluation servers, install with eval support.
• mariadb (If installing locally, install the MariaDB Connector/C first.)

Examples

uv pip install ".[mariadb,protobuf,eval]"
podman build -t $IMAGE_NAME --build-arg EXTRAS="mariadb,protobuf,eval" .

🏃Running 🏃‍♀️

Locally

uv run uvicorn src.main --host 0.0.0.0 --port 8080

Container

podman run -t $IMAGE_NAME -p 8080:8080 .

🧪 Testing 🧪

Running All Tests

To run all tests in the project:

python -m pytest

Or with more verbose output:

python -m pytest -v

Running with Coverage

To run tests with coverage reporting:

python -m pytest --cov=src

---

🔄 Protobuf Support 🔄

To process model inference data from ModelMesh models, you can install protobuf support. Otherwise, only KServe models will be supported.

Generating Protobuf Code

After installing dependencies, generate Python code from the protobuf definitions:

# From the project root
bash scripts/generate_protos.sh

Testing Protobuf Functionality

Run the tests for the protobuf implementation:

# From the project root
python -m pytest tests/service/data/test_modelmesh_parser.py -v

---

☎️ API ☎️

When the service is running, visit localhost:8080/docs to see the OpenAPI documentation!