Docker Agent、Docker Model Runner、Skillを使ったニュースまとめの作成

FROM --platform=$BUILDPLATFORM docker/docker-agent:1.32.5 AS coding-agent

FROM --platform=$BUILDPLATFORM ubuntu:22.04 AS base

LABEL maintainer="@k33g_org"
ARG TARGETOS
ARG TARGETARCH

ARG USER_NAME=docker-agent-user

ARG DEBIAN_FRONTEND=noninteractive

ENV LANG=en_US.UTF-8
ENV LANGUAGE=en_US.UTF-8
ENV LC_COLLATE=C
ENV LC_CTYPE=en_US.UTF-8

# ------------------------------------
# Install Tools
#------------------------------------
RUN <<EOF
apt-get update
apt-get install -y wget curl
apt-get clean autoclean
apt-get autoremove --yes
rm -rf /var/lib/{apt,dpkg,cache,log}/
EOF

# ------------------------------------
# Install docker-agent
# ------------------------------------
COPY --from=coding-agent /docker-agent /usr/local/bin/docker-agent

# ------------------------------------
# Create a new user
# ------------------------------------
RUN adduser ${USER_NAME}
# Set the working directory
WORKDIR /home/${USER_NAME}
# Set the user as the owner of the working directory
RUN chown -R ${USER_NAME}:${USER_NAME} /home/${USER_NAME}
# Switch to the regular user
USER ${USER_NAME}

agents:
  root:

    model: brain
    description: News Roundup Expert
    skills: true
    instruction: |
      You are an expert IT journalist with deep knowledge of software engineering, cloud infrastructure, artificial intelligence, cybersecurity, and the open-source ecosystem.
      Your role is to gather, analyze, and summarize the latest technology news in a clear, accurate, and engaging way.
      You write for a technical audience and always provide context, highlight trends, and explain the impact of each piece of news.

    toolsets:
      - type: script
        shell:

          execute_command:
            description: Execute a shell command and return its stdout and stderr output.
            args:
              command:
                description: The shell command to execute.
            cmd: |
              bash -c "$command" 2>&1

models:

  brain:
    provider: dmr
    model: huggingface.co/unsloth/qwen3.5-4b-gguf:Q4_K_M
    temperature: 0.0
    top_p: 0.95
    presence_penalty: 1.5
    provider_opts:
      # llama.cpp flags
      runtime_flags: ["--context_size=65536"]

├── .agents
│   └── skills
│       └── news-roundup
│           └── SKILL.md

---
name: news-roundup
description: search the news using Brave News Search API with a query as argument. Use this skill when the user asks to search for recent news or current events.
---
# News Roundup

## Purpose

Generate a comprehensive Markdown news report on a given topic (default: "small ai local models").

## Steps to follow

### Step 1 — Search for recent news

#### Command to execute

```bash
curl -s "https://api.search.brave.com/res/v1/news/search?q=$(echo "$ARGUMENTS_REST" | sed 's/ /+/g')&count=3&freshness=pw" \
  -H "X-Subscription-Token: ${BRAVE}" \
  -H "Accept: application/json"
```

### Step 2 — Enrich each article

For each article returned in Step 1, use the below command with the article URL to retrieve additional context and details.

#### Command to execute

```bash
curl -s "https://api.search.brave.com/res/v1/web/search?q=$(echo "$ARTICLE_URL" | sed 's/ /+/g')&count=10" \
  -H "X-Subscription-Token: ${BRAVE}" \
  -H "Accept: application/json"
```

### Step 3 — Generate the Markdown report

Using all the collected information, write a well-structured Markdown report saved to `/workspace/news-report.md`.

The report must follow this structure:

```markdown
# IT News Report — {topic}

> Generated on {date}

## Summary

A short paragraph summarizing the main trends found across all articles.

## Articles

### {Article Title}

- **Source**: {source name}
- **URL**: {url}
- **Published**: {date}

{2-3 sentence summary of the article content and its significance for IT professionals}

---

(repeat for each article)

## Key Trends

A bullet list of the main technology trends identified across all articles.
```

Save the final report to `/workspace/data/news-report-{YYYYMMDD-HHMMSS}.md` using the `write_file` tool, where `{YYYYMMDD-HHMMSS}` is the current date and time (e.g. `news-report-20260318-143012.md`).
To get the current timestamp, run:

```bash
date +"%Y%m%d-%H%M%S"
```

services:
  news-roundup:
    build:
      context: .
      dockerfile: Dockerfile
    stdin_open: true
    tty: true
    command: docker-agent run /workspace/config.yaml
    volumes:
      - ./config.yaml:/workspace/config.yaml:ro
      - ./.agents:/workspace/.agents:ro
      - ./data:/workspace/data
    working_dir: /workspace

    env_file:
      - .env

    models:
      qwen3.5-4b:

models:

  qwen3.5-4b:
    model: huggingface.co/unsloth/qwen3.5-4b-gguf:Q4_K_M
    context_size: 65536

docker compose run --rm --build news-roundup

use news roundup skill with tiny language models