[DavidTraina]

What

Adds an optional serviceName setting to the plugin config (config file or LANGFUSE_SERVICE_NAME) and makes the tracer provider honor the standard OpenTelemetry resource environment variables. The service.name resource attribute on exported spans is resolved in this order:

1. serviceName from the plugin config (or LANGFUSE_SERVICE_NAME)
2. OTEL_SERVICE_NAME / OTEL_RESOURCE_ATTRIBUTES
3. The OpenTelemetry default (unknown_service:node), unchanged from today

Why

The plugin constructs a bare NodeTracerProvider, which does not read OTEL_SERVICE_NAME or OTEL_RESOURCE_ATTRIBUTES on its own, so every span was exported as unknown_service:node with no way to change it. service.name identifies the client emitting the telemetry, so both mechanisms are useful: the config option lets users name this plugin's traces explicitly (for example to tell OpenCode traffic apart from other clients reporting to the same Langfuse project), while the environment variables keep the standard OTel way of controlling resource attributes per deployment working as users would expect. An explicit config value wins over the environment; other resource attributes from OTEL_RESOURCE_ATTRIBUTES still apply either way.

How tested

Extended the integration suite with a test that runs the built plugin against a local OTLP collector and asserts all three tiers: no configuration (default unknown_service prefix, no extra attributes), environment variables only, and config overriding OTEL_SERVICE_NAME while OTEL_RESOURCE_ATTRIBUTES values still apply. pnpm run test:integration (14 tests) and pnpm run format:check pass.

[CLAassistant]

All committers have signed the CLA.

[bezbac]

I have one minor nitpick, which I can also change myself. Otherwise, I’m very happy to merge this. Feel free to open the PR as you see fit.