I'm building an automation system for support ops, what stack makes sense?

For a support-ops automation system, the “right” stack depends mostly on volume, speed of iteration, and how many systems you need to orchestrate. A good default is to optimize for:

1. Fast integration with support tools
2. Reliable background job execution
3. Clear auditability / retries / idempotency
4. A workflow layer for human-in-the-loop steps
5. Good observability

A practical default stack

Core app

• Backend: Python FastAPI or Node.js NestJS
  • Pick FastAPI if your team likes Python and you may use AI/ML/LLM tooling.
  • Pick NestJS if your team is strongly TypeScript/JS and wants a more structured backend.

Workflow / orchestration

• Temporal
  • Best choice if the automation will have:
    • multi-step workflows
    • retries/timeouts
    • waiting on humans
    • long-running processes
    • complex branching
  • If you want something lighter to start with: Celery + Redis (Python) or BullMQ + Redis (Node), but those are less robust for business workflows than Temporal.

Database

• PostgreSQL
  • Store:
    • tickets/cases
    • workflow state
    • configs
    • audit logs
    • task history
  • Add:
    • pgvector only if you’re doing retrieval/search over knowledge bases or ticket history.

Queue / async jobs

• If using Temporal, you may not need a separate queue for workflow execution.
• If you need lightweight background jobs:
  • Redis
  • BullMQ (Node) or Celery/RQ (Python)

Integrations

• Build adapters for:
  • Zendesk / Intercom / Salesforce / Jira / Slack / email
• Use a small internal “connector” layer so each external system is isolated.

Frontend

• Next.js
  • For:
    • admin console
    • workflow builder
    • manual review queue
    • case timeline / audit UI

Auth / permissions

• Auth0, Clerk, or Keycloak depending on your needs
• Support ops usually needs:
  • SSO
  • role-based access
  • audit trails
  • per-team permissions

Observability

• OpenTelemetry
• Datadog or Grafana + Prometheus + Loki
• Strongly recommended:
  • structured logs
  • trace correlation by ticket/workflow ID
  • dead-letter / failed workflow dashboards

Deployment

• Start with:
  • Docker
  • AWS ECS/Fargate or Kubernetes if you already have platform maturity
• If early-stage, ECS/Fargate is usually simpler than Kubernetes.

---

Recommended architecture pattern

A clean pattern is:

• API service
  • receives events/webhooks
  • exposes admin APIs
• Workflow engine
  • owns all automation state
• Integration workers
  • call external systems
• Postgres
  • source of truth
• Event log
  • for replay/audit/debugging

This lets you keep support automations:

• resilient to failures
• easy to retry
• easy to inspect

---

If you want a lean MVP stack

If you’re moving fast and the system is not too complex:

• Python + FastAPI
• Postgres
• Redis
• Celery
• Next.js
• Docker
• Sentry + OpenTelemetry

This is usually enough for:

• ticket triage
• tagging/classification
• routing
• Slack notifications
• simple approvals
• KB lookup
• outbound actions

---

If you expect complex workflows soon

Go straight to:

• NestJS or FastAPI
• Temporal
• Postgres
• Next.js
• OpenTelemetry
• Datadog/Grafana
• Docker + ECS/K8s

This is better if you need:

• approvals
• SLAs
• escalations
• multi-system updates
• compensation logic
• long-running automations

---

What I’d choose by team profile

If your team is Python-heavy

• FastAPI
• Temporal (Python SDK)
• Postgres
• Next.js
• Redis only if needed

If your team is TypeScript-heavy

• NestJS
• Temporal (TypeScript SDK)
• Postgres
• Next.js
• Redis only if needed

If this is truly automation-first and mission critical

• Temporal + Postgres should be non-negotiable

---

My blunt recommendation

For a support ops automation platform, I’d start with:

FastAPI + Temporal + Postgres + Next.js + OpenTelemetry + Docker

That gives you a strong foundation without overengineering too much.

If you want, I can also give you:

1. a minimal MVP stack
2. a production-grade reference architecture
3. a buy vs build breakdown
4. a stack recommendation based on your current tools (Zendesk/Intercom/Salesforce/etc.)