Semconv Graph

Turn existing OpenTelemetry traces into a live typed entity graph without changing application instrumentation.

OTLP traces → Collector evidence → Kafka OTLP Protobuf → Java Flink lifecycle
            → schema-3 graph events → ArangoDB → typed Gremlin

The generated semantic registry keeps entity identity, relationship topology, Collector dimensions, Java extraction, Python models, and Arango collections aligned.

Runtime guarantees

• Trace-affine routing keeps both sides of an interaction on one service-graph backend.
• Positive request and discovery sums are freshness evidence; their magnitudes are discarded.
• Flink merges independent contributors and owns 24-hour freshness by default.
• Unchanged evidence refreshes expiry without repeated graph upserts.
• Nodes and relationships use deterministic IDs and identical lifecycle rules.
• Schema-3 events are complete replacements or deletes keyed by element ID.
• The indexer commits Kafka offsets only after successful Arango writes.
• Typed Gremlin reconstructs generated node and edge models without counter metadata.

Start here

• Run the focused local environment
• Read the detailed system flow
• Understand the product boundary
• Add a custom entity
• Deploy to Kubernetes
• Plan a clean upgrade