[BUG] Python OTLP exporter ignores OTEL_EXPORTER_OTLP_PROTOCOL, always uses gRPC

[shmuelarditi]

Bug Description

The tracing setup in kagent-core (tracing/_utils.py) hardcodes gRPC OTLP exporters at import time:

from opentelemetry.exporter.otlp.proto.grpc.trace_exporter import OTLPSpanExporter
from opentelemetry.exporter.otlp.proto.grpc._log_exporter import OTLPLogExporter

The OTEL_EXPORTER_OTLP_TRACES_PROTOCOL / OTEL_EXPORTER_OTLP_PROTOCOL env vars are never read, so agents cannot use http/protobuf. This blocks integration with OTLP backends that only support HTTP (e.g., Langfuse).

Steps to Reproduce

1. Set otel.tracing.exporter.otlp.protocol: http/protobuf in Helm values
2. Confirm agent pods receive OTEL_EXPORTER_OTLP_TRACES_PROTOCOL=http/protobuf
3. Agent still uses gRPC and fails:

grpc._channel._InactiveRpcError: UNKNOWN:Error received from peer {grpc_status:2, grpc_message:"Received http2 header with status: 464"}

Expected Behavior

The exporter should respect the protocol env var per the OpenTelemetry spec and select proto.http or proto.grpc accordingly.

Fix

PR #1681 addresses this with a minimal, backwards-compatible change.

Environment

• kagent: v0.8.6 (also verified on main)
• Backend: Langfuse Cloud (HTTP-only OTLP)

[boriscosic]

Just run into same issue, thanks @shmuelarditi