https://minhhh.github.io/tags/agent/Recent content in Agent on Huy Minh HaHugo -- gohugo.ioen-usMon, 08 Jun 2026 00:00:00 +0700https://minhhh.github.io/posts/pelican-to-hugo-llm/Mon, 08 Jun 2026 00:00:00 +0700https://minhhh.github.io/posts/pelican-to-hugo-llm/<p>I&rsquo;ve been running this blog on Pelican 3.4 since 2014. It worked, but the tooling was showing its age. I wanted to switch to Hugo — faster builds, simpler dependency story, and the Stack theme looked great out of the box.</p> <p>Rather than doing the migration manually, I used an AI agent (opencode) to walk through it step by step.</p> <h2 id="what-the-ai-agent-handled">What the AI agent handled </h2><ul> <li><strong>Conversion script</strong>: Wrote a Python script to convert 65 Pelican-format articles to Hugo frontmatter, handling field mapping, <code>{filename}</code> cleanup, and URL preservation.</li> <li><strong>Config scaffolding</strong>: Set up Hugo modules, the Stack theme, and all 6 config files (<code>config.toml</code>, <code>params.toml</code>, <code>menu.toml</code>, <code>permalinks.toml</code>, <code>markup.toml</code>, <code>module.toml</code>).</li> <li><strong>Deployment</strong>: Created a GitHub Actions workflow with hash-pinned actions for supply-chain security.</li> <li><strong>Asset migration</strong>: Moved avatar, favicon, and static images; verified all external image URLs.</li> <li><strong>Site verification</strong>: Ran through every page type, RSS, Disqus, syntax highlighting, dark mode, and search — all confirmed working.</li> <li><strong>Cleanup</strong>: Removed Pelican tooling (<code>fabfile.py</code>, <code>develop_server.sh</code>, <code>pelicanconf.py</code>, old themes, <code>Pipfile</code>), updated <code>Makefile</code> and README.</li> </ul> <h2 id="what-i-did">What I did </h2><ul> <li>Made decisions on menu order, category merges (Web Development → Programming), and deployment strategy (GitHub Actions vs. manual).</li> <li>Replaced the placeholder avatar with my actual photo.</li> <li>Reviewed and approved each step.</li> </ul> <h2 id="result">Result </h2><p>65 posts converted, 6 config files scaffolded, all old URLs preserved, publishing reduced to a single push:</p> <div class="highlight"><div class="chroma"> <table class="lntable"><tr><td class="lntd"> <pre tabindex="0" class="chroma"><code><span class="lnt">1 </span></code></pre></td> <td class="lntd"> <pre tabindex="0" class="chroma"><code class="language-bash" data-lang="bash"><span class="line"><span class="cl">git push origin <span class="nb">source</span> </span></span></code></pre></td></tr></table> </div> </div><h2 id="conclusion">Conclusion </h2><p>Using an AI agent for structured migration work is surprisingly effective. The heavy lifting — script generation, config wiring, verification — maps well to what agents handle best. Validation is inherent to the tools themselves: <code>hugo</code> returns build errors, <code>git diff</code> shows exactly what changed, and preview servers confirm layout. Each step self-validates, letting the agent progress without needing manual checkpoints.</p>