Fromental
A durable replacement for a large n8n research workflow, designed so every organisation is checkpointed independently and one failure can never waste an entire monthly run.
How the work moves
Client
Fromental
Agenryx system
AI & Automation
Operating change
Each organisation becomes its own retryable Postgres job
~172
Organisations in a full monthly run
44
Nodes in the workflow being replaced
1:1
One organisation to one durable job
01 · The challenge
What the client needed to resolve
The original 44-node workflow processed roughly 172 organisations inside one execution. It timed out, accumulated large execution state and behaved all-or-nothing: a late failure could waste all the successful research before it.
The change
Before
The original 44-node workflow processed roughly 172 organisations inside one execution. It timed out, accumulated large execution state and behaved all-or-nothing: a late failure could waste all the successful research before it.
Agenryx built
A FastAPI, Next.js and Postgres monorepo architecture with a dashboard trigger, per-organisation job state, Pipedrive filtering, resilient research workers, grouped report rendering and link-only email delivery through the existing n8n Gmail step.
Resolved
Each organisation becomes its own retryable Postgres job
02 · What we built
PostgreSQL acts as both database and queue. Workers claim jobs atomically with FOR UPDATE SKIP LOCKED, a reaper returns abandoned work to pending, and bounded concurrency protects external research APIs. Reports remain live web pages rather than attachments.
A FastAPI, Next.js and Postgres monorepo architecture with a dashboard trigger, per-organisation job state, Pipedrive filtering, resilient research workers, grouped report rendering and link-only email delivery through the existing n8n Gmail step.
03 · Product walkthrough
Select a step or let the walkthrough advance automatically.
The original 44-node workflow processed roughly 172 organisations inside one execution. It timed out, accumulated large execution state and behaved all-or-nothing: a late failure could waste all the successful research before it.
04 · What changed
Each organisation becomes its own retryable Postgres job
Worker crashes and redeploys recover stale work instead of restarting the batch
Salespeople receive unguessable links scoped to only their own live reports
Technology used
Start with what you need