Testbench
Kubernetes-native agent evaluation system that executes test datasets via the A2A protocol, scores responses with pluggable metrics (RAGAS by default), and publishes scores via OpenTelemetry.
📖 Documentation: https://docs.agentic-layer.ai/testbench/
Run standalone
For evaluating an agent without deploying into Kubernetes / Testkube:
pip install agentic-layer-testbench
testworkflow config.yaml
See config.example.yaml for the available configuration options.
Development
Prerequisites
- Python
- uv
- Tilt and a local Kubernetes cluster (e.g. kind)
- Testkube CLI
GOOGLE_API_KEYfor LLM-as-a-judge evaluation via Gemini
Build and run locally
# Install Python dependencies
uv sync
# Provide the LLM-as-a-judge API key
echo "GOOGLE_API_KEY=<key>" > .env
# Start the local stack (AI gateway, OTLP collector, sample agents, Testkube)
tilt up
Test
uv run poe ruff # format and lint
uv run poe mypy # static type checking
uv run poe bandit # security scanning
uv run poe test # unit tests
uv run poe check # all of the above
uv run poe test_e2e # E2E tests (requires `tilt up`)
E2E defaults target the Tilt environment. Override with E2E_DATASET_URL, E2E_AGENT_URL, E2E_MODEL if needed.
Verify the local deploy
Run the example workflow against the sample weather agent:
kubectl testkube run tw example-workflow --watch
The full walkthrough — defining experiments, configuring metrics, viewing reports — is in the first-workflow how-to.
Contributing
See the Contribution Guide.
Metadata
Release files for agentic-layer-testbench 0.10.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| agentic_layer_testbench-0.10.0.tar.gz | 38.2 kB | With provenance | Details for agentic_layer_testbench-0.10.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| agentic_layer_testbench-0.10.0-py3-none-any.whl | Python 3 | none | any | Details for agentic_layer_testbench-0.10.0-py3-none-any.whl |
Total release size: 88.2 kB
Release files / agentic_layer_testbench-0.10.0.tar.gz
| Download URL | agentic_layer_testbench-0.10.0.tar.gz |
|---|---|
| Size | 38.2 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
11ef66fb52e0dc4ad549aa9a6f0a8416e96dbd47867b4db0946cbd600744b867
|
|
BLAKE2b-256 checksum How to use checksums |
b0bd92fc3055ae3ec52aaa3282fb82f09fa15cc88f41d4c4d13c6f6edd063bdf
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.12
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jun 12, 2026.
Transparency logRelease files / agentic_layer_testbench-0.10.0-py3-none-any.whl
| Download URL | agentic_layer_testbench-0.10.0-py3-none-any.whl |
|---|---|
| Size | 50.0 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
a4846e751a0033573a4c66b5e0e1311ea105b14a3a2425626449a3abbef2a020
|
|
BLAKE2b-256 checksum How to use checksums |
8c19d3cb6c6fc2ce374d9d859b71b6a62ec9a62e1134bca61f24fa33191ee48c
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.12
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jun 12, 2026.
Transparency logSecurity for agentic-layer-testbench 0.10.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
Full provenance
Both files in this release have been uploaded with provenance metadata.
Attestation for agentic_layer_testbench-0.10.0.tar.gz
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jun 12, 2026.
Transparency logA note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.