ADK Agent Container Image

Author: george-kalisse-sada

perfkitbenchmarker/data/docker/agentic/adk-agent/.dockerignore

@@ -0,0 +1,165 @@
+
+### Python ###
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+env/
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+*.egg-info/
+.installed.cfg
+*.egg
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*,cover
+.hypothesis/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# pyenv
+.python-version
+
+# celery beat schedule file
+celerybeat-schedule
+
+# dotenv
+.env
+
+# virtualenv
+.venv/
+venv/
+ENV/
+
+# Spyder project settings
+.spyderproject
+
+# Rope project settings
+.ropeproject
+
+
+### Linux ###
+*~
+
+# temporary files which can be created if a process still has a handle open of a deleted file
+.fuse_hidden*
+
+# KDE directory preferences
+.directory
+
+# Linux trash folder which might appear on any partition or disk
+.Trash-*
+
+# .nfs files are created when an open file is removed but is still being accessed
+.nfs*
+
+
+### OSX ###
+*.DS_Store
+.AppleDouble
+.LSOverride
+
+# Icon must end with two \r
+Icon
+# Thumbnails
+._*
+# Files that might appear in the root of a volume
+.DocumentRevisions-V100
+.fseventsd
+.Spotlight-V100
+.TemporaryItems
+.Trashes
+.VolumeIcon.icns
+.com.apple.timemachine.donotpresent
+# Directories potentially created on remote AFP share
+.AppleDB
+.AppleDesktop
+Network Trash Folder
+Temporary Items
+.apdisk
+
+
+### Windows ###
+# Windows image file caches
+Thumbs.db
+ehthumbs.db
+
+# Folder config file
+Desktop.ini
+
+# Recycle Bin used on file shares
+$RECYCLE.BIN/
+
+# Windows Installer files
+*.cab
+*.msi
+*.msm
+*.msp
+
+# Windows shortcuts
+*.lnk
+
+
+### Vagrant ###
+.vagrant/
+### Local rules, see .gitignore.tail to override! ###
+shippable
+.git
+
+tmp/
+sessions.db
+.adk/

perfkitbenchmarker/data/docker/agentic/adk-agent/.gcloudignore

@@ -0,0 +1,25 @@
+# This file tells gcloud builds submit which files to exclude from the upload.
+# Without it, gcloud ignores .dockerignore and uploads everything (including .venv).
+
+.git
+.venv/
+venv/
+ENV/
+__pycache__/
+*.py[cod]
+*$py.class
+*.so
+*.egg-info/
+*.egg
+dist/
+build/
+.tox/
+.cache/
+.coverage
+htmlcov/
+*.log
+.env
+.adk/
+sessions.db
+tmp/
+.DS_Store

perfkitbenchmarker/data/docker/agentic/adk-agent/API_DOCS.md

@@ -0,0 +1,108 @@
+# Kubernetes Benchmark Agent API
+
+## Endpoints
+
+### `GET /healthz`
+Liveness probe.
+**Response:**
+```json
+{
+  "status": "ok"
+}
+```
+
+### `POST /benchmark/python/density`
+Runs the Python density benchmark.
+**Request Schema:**
+* `sample_count` (integer): Sample count per sandbox session.
+* `sample_warmup` (integer): Warmup iterations per sandbox session.
+* `concurrent_sessions` (integer): Number of parallel sandbox sessions.
+* `sandbox_exec_timeout_s` (integer): Sandbox command execution timeout in seconds.
+
+**Request Example:**
+```json
+{
+  "sample_count": 100,
+  "sample_warmup": 5,
+  "concurrent_sessions": 1,
+  "sandbox_exec_timeout_s": 60
+}
+```
+**Response Schema:**
+Returns aggregate metrics and a list of session results containing `orchestrator_cel`, `sandbox_total_cel`, and RSS memory metrics.
+
+### `POST /benchmark/python/payload`
+Runs the payload transfer benchmark.
+**Request Schema:**
+* `payload_size_mb` (float): Payload size in MB.
+* `payload_iterations` (integer): Number of transfer iterations.
+* `concurrent_sessions` (integer): Number of parallel sandbox sessions.
+* `sandbox_exec_timeout_s` (integer): Sandbox command execution timeout in seconds.
+
+**Request Example:**
+```json
+{
+  "payload_size_mb": 1.0,
+  "payload_iterations": 20,
+  "concurrent_sessions": 1,
+  "sandbox_exec_timeout_s": 60
+}
+```
+**Response Schema:**
+Returns aggregate metrics and a list of session results containing `sandbox_transfer_time_mean_ms` and throughput metrics.
+
+### `POST /benchmark/python/qps`
+Runs the QPS saturation benchmark.
+**Request Schema:**
+* `target_qps` (float): Target requests per second.
+* `duration_s` (float): Duration of the QPS burst in seconds.
+* `sandbox_exec_timeout_s` (integer): Sandbox command execution timeout in seconds.
+
+**Request Example:**
+```json
+{
+  "target_qps": 10.0,
+  "duration_s": 60.0,
+  "sandbox_exec_timeout_s": 30
+}
+```
+**Response Schema:**
+Returns `actual_qps` achieved and aggregate metrics for `ttfe_ms` and `claim_ms`.
+
+### `POST /benchmark/chromium/density`
+Runs the Chromium density benchmark.
+**Request Schema:**
+* `task_count` (integer): Iterations per Chromium session.
+* `warmup_tasks` (integer): Warmup iterations excluded from stats.
+* `concurrent_sessions` (integer): Number of parallel Chromium sessions.
+* `sandbox_exec_timeout_s` (integer): Sandbox command execution timeout in seconds.
+
+**Request Example:**
+```json
+{
+  "task_count": 10,
+  "warmup_tasks": 2,
+  "concurrent_sessions": 1,
+  "sandbox_exec_timeout_s": 120
+}
+```
+**Response Schema:**
+Returns aggregate metrics and a list of session results containing interaction, navigate, evaluate, click, and screenshot latencies.
+
+### `POST /run`
+Raw ADK agent interaction.
+**Request Schema:**
+* `prompt` (string): The prompt to send to the agent.
+
+**Request Example:**
+```json
+{
+  "prompt": "(Unused now, as agent has a MockLLM)"
+}
+```
+**Response Schema:**
+```json
+{
+  "response": "Agent output string"
+}
+```

perfkitbenchmarker/data/docker/agentic/adk-agent/Dockerfile

@@ -0,0 +1,29 @@
+FROM python:3.13-slim
+WORKDIR /app
+
+# Install kubectl (required by k8s-agent-sandbox for port-forwarding to sandbox pods)
+# Uses TARGETARCH (injected by BuildKit) to download the correct binary for amd64 or arm64
+RUN apt-get update && \
+    apt-get install -y --no-install-recommends curl ca-certificates && \
+    ARCH=$(dpkg --print-architecture) && \
+    curl -LO "https://dl.k8s.io/release/$(curl -sL https://dl.k8s.io/release/stable.txt)/bin/linux/${ARCH}/kubectl" && \
+    install -o root -g root -m 0755 kubectl /usr/local/bin/kubectl && \
+    rm kubectl && \
+    apt-get purge -y curl && \
+    apt-get autoremove -y && \
+    rm -rf /var/lib/apt/lists/*
+
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+RUN adduser --disabled-password --gecos "" myuser && \
+    chown -R myuser:myuser /app
+
+COPY . .
+
+USER myuser
+
+ENV PATH="/home/myuser/.local/bin:$PATH"
+
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8080"]
+

perfkitbenchmarker/data/docker/agentic/adk-agent/README.md

@@ -0,0 +1,56 @@
+# ADK Agent Container
+
+FastAPI service deployed inside the Kubernetes cluster that serves as the
+orchestrator for Agent Sandbox benchmarks. PKB benchmark modules
+communicate with this service via HTTP to trigger benchmark runs.
+
+## Architecture
+
+```
+PKB (host machine)                    Kubernetes Cluster
+-----------------                     -----------
+benchmark.Run()
+  -> HTTP POST /benchmark/...  -->  main.py (FastAPI)
+                                      -> Runner(agent=root_agent)
+                                        -> MockLlm yields code
+                                        -> BenchmarkExecutor
+                                          -> SandboxClient.create_sandbox()
+                                          -> sandbox.files.write("script.py", code)
+                                          -> sandbox.commands.run("python3 script.py")
+                                          -> SandboxClient.delete_sandbox()
+```
+
+## Components
+
+| File | Purpose |
+|------|---------|
+| `main.py` | FastAPI service with benchmark endpoints |
+| `performance_agent/agent.py` | ADK agent definition with MockLlm and BenchmarkExecutor |
+| `sandboxed_apps/python_test_app/benchmark_density.py` | Python density benchmark (runs inside gVisor sandbox) |
+| `sandboxed_apps/python_test_app/benchmark_payload.py` | Payload transfer benchmark (runs inside gVisor sandbox) |
+| `sandboxed_apps/python_test_app/benchmark_qps.py` | QPS validation script (runs inside gVisor sandbox) |
+
+## Endpoints
+
+| Method | Path | Description |
+|--------|------|-------------|
+| GET | `/healthz` | Liveness probe |
+| POST | `/benchmark/python/density` | Python sandbox density measurement |
+| POST | `/benchmark/python/payload` | Payload transfer measurement |
+| POST | `/benchmark/python/qps` | QPS saturation measurement |
+| POST | `/benchmark/chromium/density` | Chromium browser density measurement |
+| POST | `/run` | Raw ADK agent interaction |
+
+## Building
+
+The image is built automatically by PKB during the provision stage via
+`container_specs`. For ARM64, use the `cloudbuild-arm64.yaml` config
+with `gke_prerequisites.py --target_arch=arm64`.
+
+## Dependencies
+
+See `requirements.txt`. Key packages:
+- `google-adk[gke,extensions]` -- Google Agent Development Kit
+- `k8s-agent-sandbox` -- Sandbox lifecycle client
+- `fastapi` + `uvicorn` -- HTTP service
+- `playwright` -- Chromium CDP automation (orchestrator-side)

perfkitbenchmarker/data/docker/agentic/adk-agent/__init__.py

@@ -0,0 +1 @@
+# ADK Agent package