spalmurray-codecov
commented
10 hours ago @spalmurray-codecov @ajay-sentry Nov 18th meeting notes:
- Part one add SDK, try firing some events
- Make slack channel
- we should make a strategy pattern with standard payload
- Perk: when we add new events it's easy, as it should be
- without a framework this gets hairy
- What do our payloads look like for customer, owner, button, etc.
- payloads give a snapshot of a resource at a given time, allows to build a picture over time
- works well with amplitude's system for visualizations
- on customer updated model, have customer *status* which is 1/8 steps for example (file uploaded, oauthed, etc.)
- allows to filter customers by what step they're on
- First class status on events to drive funnel
- Extra attributes allow for further filtering
- once we have SDKs built out, work closely with product to figure out what we care about (owner? plan?), base off what they say we investigate our context to build a generalized schema for important things that are always available.
- On each event, we have this standard context on each model.
- Maybe to constrain this for the quarter maybe focus on one event, one schema, and fire them.
- exact API:
- fire or publish
- two params, event name, strictly typed payload
- validate at runtime? build time? sanitization?
- firing events on the frontend you will have browser info, tz info -> system time, FE specific payload fields?
- base values vs extra metadata that's platform specific.
- Maybe proxy through backend to get more context.
- Maybe we don't even need python SDK this Q? maybe not JS sdk?
- We'll see how this week goes, but aim for Spencer having a doc together to review with the gang in person in YYZ.
- Bring up with Katia and AJ scope change on this. Initial plan was much smaller.
- Maybe this Q becomes only basic integration with planning as a side project until next Q.```
Settle on a schema to use for our Codecov Amplitude metrics using the initial Onboarding events as a guide. Keep in mind we want this to be generic and work across product features and applications.