The latest articles on DEV Community by Arya Mehta (@arya_mehta_26). https://dev.to/arya_mehta_26 https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3913616%2F7ea55c83-1954-415a-89cc-743815a652e3.gif https://dev.to/arya_mehta_26 en Arya Mehta Tue, 05 May 2026 09:30:15 +0000 https://dev.to/arya_mehta_26/building-a-multi-step-ai-pipeline-workspace-2-oss-fixes-to-rocketride-server-mho https://dev.to/arya_mehta_26/building-a-multi-step-ai-pipeline-workspace-2-oss-fixes-to-rocketride-server-mho <p>I’ve been learning RocketRide (open-source AI pipeline platform + VS Code extension) by building a small project and contributing fixes upstream.</p> <h2> What RocketRide is (quick) </h2> <p>RocketRide lets you build AI workflows as <strong><code>.pipe</code></strong> graphs inside VS Code (visual canvas), run them against a local/hosted engine (default settings often point to <code>http://127.0.0.1:5565</code>), and inspect step-level execution/traces.</p> <h2> What I built: a pipeline workspace (Next.js) </h2> <p>I shipped a product-shaped demo that mirrors the “runs → steps → traces → artifact” experience:</p> <ul> <li> <strong><code>/</code></strong>: landing page + quick explanation</li> <li> <strong><code>/app</code></strong>: enter a goal/prompt → run a multi-step pipeline → see: <ul> <li>step-by-step execution trace (detail + timing)</li> <li>a final <strong>markdown artifact</strong> (structured handoff doc)</li> <li>recent runs + seeded demo rows (useful in fresh Docker/CI environments)</li> </ul> </li> </ul> <p>Repo: <a href="https://github.com/aryaMehta26/multi-step-ai-pipelines" rel="noopener noreferrer">https://github.com/aryaMehta26/multi-step-ai-pipelines</a></p> <h3> Reproducibility (important for reviewers) </h3> <ul> <li> <strong>Docker</strong>: <code>docker compose up --build</code> → <a href="http://localhost:3000" rel="noopener noreferrer">http://localhost:3000</a> </li> <li> <strong>CI</strong>: GitHub Actions runs <code>npm ci → lint → build</code> </li> </ul> <h2> “Built on RocketRide” evidence </h2> <p>Alongside the web app, I included an actual RocketRide pipeline file + setup guide:</p> <ul> <li> <code>.pipe</code>: <code>pipelines/goal-planning-agent.pipe</code> </li> <li>docs: <code>docs/ROCKETRIDE.md</code> </li> </ul> <p>So the submission shows both:<br> 1) RocketRide-native <code>.pipe</code> workflow in the extension<br> 2) a web UI showing how I’d present pipeline execution/traces outside the IDE</p> <h2> OSS contributions (2 upstream PRs) </h2> <p>While building, I fixed a couple small-but-real edge cases upstream:</p> <ul> <li><p>PR #754 (test mocks): fixes Python truthiness traps (<code>bool</code> treated as <code>int</code>, <code>limit=0</code>, <code>0.0</code> sorting)<br> <a href="https://github.com/rocketride-org/rocketride-server/pull/754" rel="noopener noreferrer">https://github.com/rocketride-org/rocketride-server/pull/754</a></p></li> <li><p>PR #755 (Python SDK): fixes websocket URI normalization so it doesn’t create <code>//task/service</code> or duplicate <code>/task/service</code><br> <a href="https://github.com/rocketride-org/rocketride-server/pull/755" rel="noopener noreferrer">https://github.com/rocketride-org/rocketride-server/pull/755</a></p></li> </ul> <p>Fork: <a href="https://github.com/aryaMehta26/rocketride-server" rel="noopener noreferrer">https://github.com/aryaMehta26/rocketride-server</a></p> <h2> Feedback welcome </h2> <p>If you’ve built workflow UIs / agentic pipelines before, I’d love feedback on:</p> <ul> <li>what makes traces actually useful (beyond “logs everywhere”)</li> <li>what you’d want in a pipeline runner UI for non-IDE users</li> </ul> <p>Thanks for reading!</p> ai nextjs opensource showdev