Add Docker support for Mermaid benchmark tooling

Dockerfile

@@ -0,0 +1,50 @@
+# syntax=docker/dockerfile:1
+FROM ghcr.io/astral-sh/uv:python3.13-bookworm
+
+# Install Node.js, npm, and Chromium dependencies for mermaid-cli
+RUN apt-get update \
+    && apt-get install -y --no-install-recommends \
+        fonts-liberation \
+        fonts-noto-color-emoji \
+        libasound2 \
+        libatk-bridge2.0-0 \
+        libatk1.0-0 \
+        libatspi2.0-0 \
+        libcairo2 \
+        libcups2 \
+        libdrm2 \
+        libgbm1 \
+        libgtk-3-0 \
+        libnss3 \
+        libpango-1.0-0 \
+        libx11-xcb1 \
+        libxcomposite1 \
+        libxdamage1 \
+        libxfixes3 \
+        libxi6 \
+        libxrandr2 \
+        libxrender1 \
+        libxshmfence1 \
+        libxss1 \
+        libxtst6 \
+        nodejs \
+        npm \
+    && rm -rf /var/lib/apt/lists/*
+
+# Install mermaid CLI globally
+RUN npm install -g @mermaid-js/mermaid-cli
+
+# Set work directory
+WORKDIR /app
+
+# Copy dependency files first for better caching
+COPY pyproject.toml uv.lock ./
+
+# Copy source
+COPY . .
+
+# Sync dependencies during build so they are baked into the image
+RUN uv sync --frozen
+
+# Default entrypoint
+ENTRYPOINT ["/app/docker/merbench-entrypoint.sh"]

Makefile

@@ -15,4 +15,14 @@ adk_basic_ui:
 	uv run adk web agents_mcp_usage/basic_mcp
 
 adk_multi_ui:
-	uv run adk web agents_mcp_usage/multi_mcp
+	uv run adk web agents_mcp_usage/multi_mcp
+
+merbench-docker-build:
+	docker build -t merbench .
+
+merbench-docker-run:
+	docker run --rm \
+	  -e GEMINI_API_KEY=$${GEMINI_API_KEY} \
+	  -e OPENAI_API_KEY=$${OPENAI_API_KEY} \
+	  -v "$$(pwd)/mermaid_eval_results:/app/mermaid_eval_results" \
+	  merbench $${ARGS}

README.md

@@ -44,6 +44,34 @@ Run an Agent framework script e.g.:
 
 Check console, Logfire, or the ADK web UI for output
 
+## Docker quickstart
+
+Build the pre-baked evaluation image (includes Python dependencies and the Mermaid CLI):
+
+```bash
+docker build -t merbench .
+```
+
+Run the multi-model Mermaid benchmark with your API keys and a bind mount so results persist on the host:
+
+```bash
+docker run --rm \
+  -e GEMINI_API_KEY="your-gemini-key" \
+  -e OPENAI_API_KEY="your-openai-key" \
+  -v "$(pwd)/mermaid_eval_results:/app/mermaid_eval_results" \
+  merbench --models gemini-1.5-pro,openai-gpt-4.1-mini
+```
+
+The container entrypoint defaults to `run_multi_evals.py`. Override it to launch other tooling, such as the evaluation UI:
+
+```bash
+docker run --rm \
+  -e GEMINI_API_KEY="your-gemini-key" \
+  -v "$(pwd)/mermaid_eval_results:/app/mermaid_eval_results" \
+  --entrypoint uv \
+  merbench run agents_mcp_usage/evaluations/mermaid_evals/merbench_ui.py
+```
+
 ## Project Overview
 
 This project aims to teach:

docker/merbench-entrypoint.sh

@@ -0,0 +1,7 @@
+#!/bin/sh
+set -e
+
+# Ensure results directory exists if bind-mounted
+mkdir -p /app/mermaid_eval_results
+
+exec uv run agents_mcp_usage/evaluations/mermaid_evals/run_multi_evals.py "$@"