[WIP] feat: add MLflow tracing integration via OpenTelemetry bridge

[luis5tb]

Summary

• Add MLflow LLM observability support via OpenTelemetry bridge — Google ADK natively generates OTel traces, MLflow (>= 3.6.0) accepts them at its /v1/traces OTLP endpoint
• MLflow tracing is independently toggleable from OTel tracing (MLFLOW_ENABLED works without OTEL_ENABLED)
• On Cloud Run, the agent connects to an external MLflow instance (Red Hat infrastructure) — no new GCP services deployed
• On OpenShift, MLflow Tracking Server is deployed as an optional Helm component with dedicated PostgreSQL, artifact PVC, NetworkPolicy, and optional Route

Changes

Agent code:

• config/settings.py: 6 new MLflow settings (enabled, tracking_uri, experiment_name, experiment_id, log_prompts, run_tags)
• telemetry/setup.py: _add_mlflow_processor() adds a second OTLP span processor targeting MLflow; setup_telemetry() restructured so MLflow works independently of otel_enabled
• pyproject.toml: new mlflow optional dependency group (mlflow-tracing>=3.6.0, opentelemetry-exporter-otlp-proto-http>=1.20.0)
• .env.example, Makefile: env var docs + lock-mlflow target

Cloud Run: env vars in service.yaml pointing at external MLflow

OpenShift: 7 new Helm templates (Deployment, Service, Route, PVC, PostgreSQL, PostgreSQL PVC, NetworkPolicy) + values.yaml, configmap.yaml, secret.yaml, _helpers.tpl updates

Tests: 20 test cases covering settings, enabled/disabled paths, independent MLflow (no OTel), experiment ID headers

Test plan

• ruff check — all checks passed
• pytest tests/ — 470 passed, 0 failures
• Deploy on OpenShift with mlflow.enabled: true and verify traces appear in MLflow UI
• Deploy on Cloud Run with MLFLOW_ENABLED=true pointing at external MLflow and verify traces
• Verify MLflow works with OTEL_ENABLED=false (independent mode)
• Verify no impact when MLFLOW_ENABLED=false (default)

🤖 Generated with Claude Code