Bryn Cooke
Bryn Cooke
Maybe we can support functions that are not on an impl ### Alternative 1: Directly on `fn` outside of impl. ``` #[builder(entry="foo", exit="call") fn do_foo(a: usize) { ... } ```...
Currently the filter layer allows processing to be halted, but an error must be returned because Result is used to indicate continued processing. It'd be good to be able to...
Prometheus metrics, once enabled are not secured. Call this our in the documentation and give suggestions to user about securing at ingress or service mesh.
Recently we introduced a `PluginTestHarness`. It's very basic and needs to work. Things that need to be fixed: 1. Everything must be mocked. If you make a request then if...
Decision below is to remove the cache layer --- Original body: --- What layers do we want to expose to plugin authors? We currently have: * checkpoint * instrument *...
There's a couple fields in our `reports.proto` that we are not populating right now that would help power the `https://studio.apollographql.com/graph/{GRAPH_ID}/operations?tab=errors&variant={VARIANT}` page that we would be good to add to the...
Just get an initial investigation done. ## Potential follow-ups - [ ] Plugin/layer config version implementation
Currently we have: * Router - RouterRequest, RouterResponse * QueryPlanning - QueryPlannerRequest, QueryPlannerResponse * Execution - ExecutionRequest, ExecutionResponse * Subgraph - SubgraphRequest, SubgraphResponse In particular the RouterService seems badly named....
Create a comprehensive list of all errors. Ensure that error messages have: 1. Information about what happened in a user friendly language. 2. Information about what to do about the...
Unlike the rest of the data-structures QueryPlannerResponse doesn't have an errors field. This is problematic as the error handling mechanism in tower is only really suitable for unrecoverable pipeline errors....