Webhooks, Listeners, and Events?

[chadptk1238]

if we wanted to build agents that are proactive for things like a inbound message or form fill. is polling the only/best option to retain Auth via Arcade?
Are you guys planning on building out any Proactive features?

[evantahler]

Thanks for starting this conversation, @chadptk1238! This is certainly a topic on our roadmap. It would be helpful if you could share more about your use case. I know this is a generic sort of ask, so here are some topics that are on my mind when I think about this feature:

• Delivery: Will your agent always be running and want us to send a webhook, or will your agent be periodically woken up and want to consume a queue of pending messages?
• Filtering: How are you thinking about filtering events? e.g. you probably don't want every email your end-user gets to trigger the event, but there's some filter (emails from a certain person or domain?) which should be trigger the event to your agent.
• SLAs: How important is "realtime" to you? As many APIs /don't/ have webhooks themselves, the Arcade platform would end up needing to poll the upstream systems, and the cost for that increases as the delay shrinks. Furthermore, polling at a high frequency requires higher rate-limits. Would you be willing to pay (salesforce/slack/google/etc) for that higher limit?
• Throttling and Batching - Say 10 events come in that match your filters - should we send you 10 unique events or one "batch" event so your agent can process them all at once? There's probably cost-savings to process the batch, but that would increase latency.
• Synthetic events - Are you interested in "generated" events, e.g. cron-like events to wake your agent up in some way "Every hour" or "once a day"?
• Hydration: - How much data do you need within the event. Consider an event from Notion saying "page 43 is updated". Would you also need the full content of page 43 (and the previous content?). Does the diff matter to you? Fully hydrating the object adds time, but might increase usefulness. Are you doing a CDC-type workflow?

If folks responding to this issue could comment on the above, that would be very helpful!