<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Vectorbea Engineering</title>
    <link>https://engineering.vectorbea.com</link>
    <atom:link href="https://engineering.vectorbea.com/rss.xml" rel="self" type="application/rss+xml" />
    <description>Engineering notes from the team building Vectorbea, a durable, long-running visual agentic workflow builder. Execution guarantees, retries, checkpoints, approvals, BYOK, and worker orchestration.</description>
    <language>en-us</language>
    
    <item>
      <title>Lessons from Building Vectorbea v1</title>
      <link>https://engineering.vectorbea.com/articles/lessons-from-building-vectorbea-v1</link>
      <guid>https://engineering.vectorbea.com/articles/lessons-from-building-vectorbea-v1</guid>
      <pubDate>Tue, 26 May 2026 00:00:00 GMT</pubDate>
      <description>What we&apos;d keep and what we&apos;d change across UI, backend, security, observability, and positioning, after shipping the first version of Vectorbea&apos;s durable workflow engine.</description>
      <author>Susmit Banerjee</author>
      <category>lessons</category>
      <category>retrospective</category>
      <category>engineering-culture</category>
    </item>

    <item>
      <title>Self-Correction Loops for Failed Workflows: Blind Retry Isn&apos;t Intelligence</title>
      <link>https://engineering.vectorbea.com/articles/self-correction-loops-for-failed-workflows</link>
      <guid>https://engineering.vectorbea.com/articles/self-correction-loops-for-failed-workflows</guid>
      <pubDate>Tue, 05 May 2026 00:00:00 GMT</pubDate>
      <description>The difference between retrying a failed step and helping a workflow understand why it failed, error classification, bounded self-correction, and where we draw the line and call a human.</description>
      <author>Susmit Banerjee</author>
      <category>agentic-systems</category>
      <category>self-correction</category>
      <category>reliability</category>
    </item>

    <item>
      <title>Cost Budgets and Rate Limits for Agentic Workflows</title>
      <link>https://engineering.vectorbea.com/articles/cost-budgets-and-rate-limits</link>
      <guid>https://engineering.vectorbea.com/articles/cost-budgets-and-rate-limits</guid>
      <pubDate>Tue, 21 Apr 2026 00:00:00 GMT</pubDate>
      <description>How we estimate token costs before and during a run, enforce per-run and per-workspace budgets, apply rate limits, and build kill switches that actually stop a runaway workflow.</description>
      <author>Susmit Banerjee</author>
      <category>cost</category>
      <category>rate-limiting</category>
      <category>reliability</category>
    </item>

    <item>
      <title>Worker Scaling with Redis Streams: Consumer Groups, PEL, and When to Reach for Kafka</title>
      <link>https://engineering.vectorbea.com/articles/worker-scaling-with-redis-streams</link>
      <guid>https://engineering.vectorbea.com/articles/worker-scaling-with-redis-streams</guid>
      <pubDate>Tue, 07 Apr 2026 00:00:00 GMT</pubDate>
      <description>How Vectorbea&apos;s worker fleet pulls work from Redis Streams, consumer groups, the pending entries list, retry and DLQ handling, and the honest answer to &apos;why not Kafka?&apos;</description>
      <author>Susmit Banerjee</author>
      <category>infrastructure</category>
      <category>redis</category>
      <category>workers</category>
      <category>scaling</category>
    </item>

    <item>
      <title>BYOK Architecture for an AI SaaS: Benefits, Risks, and Boundaries</title>
      <link>https://engineering.vectorbea.com/articles/byok-architecture-for-ai-saas</link>
      <guid>https://engineering.vectorbea.com/articles/byok-architecture-for-ai-saas</guid>
      <pubDate>Tue, 24 Mar 2026 00:00:00 GMT</pubDate>
      <description>Why we let customers bring their own LLM provider keys, what it costs them and us, and the security boundaries we think any BYOK system needs, without the implementation specifics.</description>
      <author>Susmit Banerjee</author>
      <category>security</category>
      <category>byok</category>
      <category>architecture</category>
    </item>

    <item>
      <title>Human Approval Gates in Agentic Systems</title>
      <link>https://engineering.vectorbea.com/articles/human-approval-gates-in-agentic-systems</link>
      <guid>https://engineering.vectorbea.com/articles/human-approval-gates-in-agentic-systems</guid>
      <pubDate>Mon, 09 Mar 2026 00:00:00 GMT</pubDate>
      <description>Modeling &apos;wait for a human&apos; as a first-class workflow step, waiting states, timeouts, escalation, and why the audit trail has to be airtight.</description>
      <author>Susmit Banerjee</author>
      <category>agentic-systems</category>
      <category>human-in-the-loop</category>
      <category>ux</category>
    </item>

    <item>
      <title>Retries, Resume, and Idempotency: The Unglamorous Core of Reliability</title>
      <link>https://engineering.vectorbea.com/articles/retries-resume-and-idempotency</link>
      <guid>https://engineering.vectorbea.com/articles/retries-resume-and-idempotency</guid>
      <pubDate>Wed, 18 Feb 2026 00:00:00 GMT</pubDate>
      <description>Step-level retries sound simple until you ask what happens when the step already had a side effect. Notes on idempotency keys, resumption, and duplicate tool calls in Vectorbea.</description>
      <author>Susmit Banerjee</author>
      <category>reliability</category>
      <category>idempotency</category>
      <category>retries</category>
    </item>

    <item>
      <title>Designing Event History as a Primitive for AI Workflows</title>
      <link>https://engineering.vectorbea.com/articles/designing-event-history-for-ai-workflows</link>
      <guid>https://engineering.vectorbea.com/articles/designing-event-history-for-ai-workflows</guid>
      <pubDate>Mon, 02 Feb 2026 00:00:00 GMT</pubDate>
      <description>How we modeled the append-only event log that backs every Vectorbea run, and why we treat it as the source of truth rather than an audit trail bolted on afterward.</description>
      <author>Susmit Banerjee</author>
      <category>event-sourcing</category>
      <category>architecture</category>
      <category>observability</category>
    </item>

    <item>
      <title>Why Long-Running AI Workflows Need Durable Execution</title>
      <link>https://engineering.vectorbea.com/articles/why-long-running-ai-workflows-need-durable-execution</link>
      <guid>https://engineering.vectorbea.com/articles/why-long-running-ai-workflows-need-durable-execution</guid>
      <pubDate>Mon, 12 Jan 2026 00:00:00 GMT</pubDate>
      <description>Async jobs and retry decorators get you most of the way to a working agent, and then they don&apos;t. Here&apos;s why we built Vectorbea around durable execution from day one.</description>
      <author>Susmit Banerjee</author>
      <category>durable-execution</category>
      <category>architecture</category>
      <category>reliability</category>
    </item>
  </channel>
</rss>