[Feature]: Propagate retry info in response

[Pratham-Mishra04]

Prerequisites

• I have searched existing issues and discussions to avoid duplicates

Problem to solve

Right now there is no way of knowing how many retries were done before the request was successful

Proposed solution

Pass on the number of retries attempted information in BifrostResponse and BifrostError

Alternatives considered

No response

Area(s)

Core (Go)

Additional context

No response

[sureshkrishnan-v]

Hi, can i work on this feature as a new contributer it would be great if any one guides me

[Pratham-Mishra04]

Hello @sureshkrishnan-v , yes absolutely happy to guide you here.

1. Get familiar with the repo structure and the separation between core (Go SDK) and transports (gateway built on top of core).
2. The file bifrost.go in /core contains all the core logic for fallbacks and retries.
3. ctx is the shared context, so you can pass retry information down using the shared context type schemas.BifrostContextKey.
4. For the gateway, you can capture that context in HTTP handlers and propagate it back via a header like x-bf-retries.
5. Add a Prometheus variable for retries and extend our telemetry plugin.
6. Finally, extend our documentation: update the Managing Retries section in the Provider Configuration docs of both the gateway and Go SDK, explaining how to check retry counts, add a section in the Fallbacks documentation to explain how retries are managed alongside fallbacks, and include the new Prometheus label in the telemetry docs.

Let’s keep the conversation on this thread, happy to help along the way😄

[sureshkrishnan-v]

hi @Pratham-Mishra04 here you can see i cannot use variable schemas.BifrostContextKeyRetryCount even though i have declared in schemas/bifrost.go

[Pratham-Mishra04]

yes that's cause the transports still refers to the released version of core, and not the local build. Make sure to run make dev at project root to setup your local for development :)