Automend
Visual DAG Workflow Automation Platform with Headless OAuth & Distributed Queues
01.The Problem & Why Naive Implementations Fail
Third-party workflow tools (Zapier, Make) are expensive, vendor-locked, and brittle when user OAuth credentials break when team members leave an organization.
02.Architectural Solution & Engineering Decisions
Engineered an interactive drag-and-drop workflow builder with typed triggers and parameters. Built decoupled background worker queues via BullMQ and DragonflyDB/Redis, and anchored headless OAuth bot tokens to the workspace to preserve pipeline execution after member offboarding.
Key Architectural Highlights
Interactive Flow Canvas DAG Builder
Node-edge graph canvas with typed webhook inputs, conditional branches, and parameter configuration drawers.
Distributed Worker Queues
Hono API route producers feed distributed BullMQ queues with backpressure management, automatic retries, and Bull Board observability.
Persistent Headless Bot Token Lifecycle
PKCE OAuth architecture anchoring Slack and Google bot tokens to the shared workspace package rather than individual personal accounts.