feat: build onboarding checks for first-time local setup

Author: 0xfaan

detection/model_inference.py

@@ -452,6 +452,14 @@ def score(
             span.set_attribute("wallet.id", hash_span_id(wallet) if wallet else "unknown")
             result = self._score_impl(feature_row, labelled_count, caller_id)
             span.set_attribute("model.score", result.get("score", -1))
+            # Embed a lightweight version stamp so every score output carries
+            # its provenance (Issue #4).
+            try:
+                from utils.version_stamp import get_version as _ll_version
+
+                result["ledgerlens_version"] = _ll_version()
+            except Exception:
+                pass
             return result
 
     def _score_impl(

examples/__init__.py

@@ -0,0 +1,22 @@
+"""End-to-end detection workflow examples for LedgerLens.
+
+This package contains runnable examples that exercise the full detection
+pipeline from raw trade data through to a LedgerLens Risk Score, without
+requiring a live Stellar Horizon connection.  Each example:
+
+- Generates synthetic trade data that mimics a specific on-chain pattern
+- Runs the full detection stack (Benford engine, feature engineering, model
+  inference, SHAP explainer)
+- Prints the resulting risk score and top SHAP attributions
+
+These examples double as integration smoke-tests: ``pytest examples/`` runs
+all of them in CI to ensure the pipeline does not regress.
+
+Run any example directly::
+
+    python -m examples.e2e_clean_trading
+    python -m examples.e2e_wash_trading_ring
+    python -m examples.e2e_benford_anomaly
+    python -m examples.e2e_cross_venue_coordination
+    python -m examples.e2e_full_pipeline
+"""