Testbench

Kubernetes-native agent evaluation system that executes test datasets via A2A protocol, evaluates responses using pluggable metrics (RAGAS by default), and publishes scores via OpenTelemetry.

Documentation

Full documentation is available at docs.agentic-layer.ai.

Prerequisites

• Python 3.12+ and uv
• Kubernetes cluster (e.g. kind) with Tilt
• Testkube CLI
• GOOGLE_API_KEY for LLM-as-a-judge evaluation via Gemini models

Getting Started

# 1. Start local infrastructure (AI Gateway, OTLP collector, sample agents, Testkube)
#    Create a .env file with GOOGLE_API_KEY=your-key first
tilt up

# 2. Run the example evaluation workflow
kubectl testkube run tw example-workflow --watch

See the how-to guide for detailed pipeline usage including dataset format, metric configuration, and custom workflows.

Development

Command	Description
uv run poe check	Run all quality checks (tests, mypy, bandit, ruff)
uv run poe test	Unit tests
uv run poe format	Format with Ruff
uv run poe lint	Lint and auto-fix with Ruff
uv run poe ruff	Both format and lint
uv run poe mypy	Static type checking
uv run poe bandit	Security vulnerability scanning

E2E Testing

Requires the Tilt environment running (tilt up).

# Configure (optional — defaults target the Tilt environment)
export E2E_DATASET_URL="http://data-server.data-server:8000/dataset.csv"
export E2E_AGENT_URL="http://weather-agent.sample-agents:8000"
export E2E_MODEL="gemini-2.5-flash-lite"

# Run
uv run poe test_e2e

Contributing

See Contribution Guide for details on contributing and the process for submitting pull requests.