August 2023 Summaries

Semi-Autonomous AI Agents are a middle ground between fully autonomous AI agents and human decision making. They leverage the power of Large Language Models (LLMs) to offer suggestions while still allowing users to maintain control over their actions. This approach is particularly useful in handling complex asynchronous workflows, where decisions need to be made at different stages of a process. Inngest, a platform for building complex async workflows within Next.js applications, shines in managing such workflows by providing a clear SDK and executing long-running serverless processes using existing hosting infrastructure. By combining Inngest with other tools like PartyKit, developers can create chat UIs that interact with external APIs and enable users to perform actions on suggested issues or tasks.

The article discusses sending customer lifecycle emails using Resend and Inngest. It emphasizes the importance of not annoying customers while ensuring reliable email delivery. Customer lifecycle emails are sent at specific milestones in a user's journey within an application, such as account creation or performing critical actions. To handle these events, the article suggests using durable, asynchronous event-driven workflows with Inngest and sending emails through Resend. The example provided demonstrates how to create an onboarding campaign for new users, which includes a welcome email, waiting for user activity, and sending follow-up emails based on their actions.

The article discusses building an event-driven video processing workflow using Next.js, tRPC, and Inngest. It outlines the steps involved in uploading a video, editing metadata, monitoring progress, converting media into appropriate formats for global streaming delivery, and making videos accessible with transcripts and subtitles. The author explains how they used various services like Vercel, Cloudinary, Sanity, Mux, Deepgram, and OpenAI to create this workflow. They also highlight the role of Inngest in providing consistent, reliable, resilient, and understandable workflows throughout the complex asynchronous process. The article concludes by discussing how this workflow can be further improved and expanded upon for more complex content publishing pipelines.