We're scaling our internal support ticketing. 200 users, mix of engineers and L1/L2 support. Need to ingest from Zendesk, enrich with internal API data (customer tier, SLA), trigger PagerDuty for critical bugs, and sync summaries to a Postgres analytics DB.
Currently a patchwork of scripts. Looking at Lindy and Tray.io to consolidate.
Key requirements:
* Must handle ~5k tickets/day.
* Cost control is critical. Vendor "contact for pricing" is a red flag.
* Failures must be observable. We need logs, retry queues, clear error propagation.
* Minimal vendor lock-in. Prefer to define workflows as code/in config.
Initial findings:
**Lindy**
* API-centric. Can define the entire pipeline in a YAML-ish config.
* Self-hosted agent possible (Kubernetes). Appeals to infra cost concerns.
* Example enrichment step:
```yaml
- step: enrich_customer_data
action: http.post
input: {{trigger.ticket.customer_id}}
url: {{secrets.INTERNAL_API}}/customer/tier
on_failure: retry_3x
```
* Pricing per "active workflow". Straightforward.
**Tray.io**
* GUI builder primary, CLI/API secondary.
* "Magic" data mapper. Less transparent on transformation logic.
* Built-in connectors are polished, but you pay for it.
* Their sales team is pushing "unlimited tasks" but requires a 50-user minimum commit.
Biggest concern with Tray is opacity. When a mapping fails, debugging is a black box. Lindy's config-atleast shows the pipeline structure.
Anyone run both at scale for a similar use-case? Specifically:
* Real-world cost per 5k daily operations?
* Observability integration (Prometheus, Grafana)?
* Failure modes and incident response stories?
latency kills