feat(batch-agent): add journey eval to E2E harness

- journey_runner.py: orchestrates journey start → simulated user
  messages → template extraction → LLM judge scoring
- config.py: JourneyFixture dataclass with user_messages and
  expected_template_criteria, discover_journey_fixtures()
- langfuse_eval.py: sync_journey_fixture_to_dataset()
- cli.py: new 'journey' subcommand (python -m eval journey)
  with --fixture, --models, --judge-model flags
- fixtures/journey_invoice_setup.yaml: example journey fixture
  with 4 user messages and 8 quality criteria

services/batch-agent/eval/langfuse_eval.py

@@ -96,6 +96,52 @@ def sync_fixture_to_dataset(fixture: EvalFixture) -> str | None:
     return dataset_name
 
 
+def sync_journey_fixture_to_dataset(fixture) -> str | None:
+    """Create or update a Langfuse dataset from a journey fixture.
+
+    Each journey fixture becomes a single dataset item with:
+    - input: {directory, data_types, user_messages}
+    - expected_output: {criteria}
+    """
+    lf = _get_langfuse()
+    if lf is None:
+        logger.info("langfuse_eval: Langfuse not configured — skipping journey dataset sync")
+        return None
+
+    dataset_name = f"journey-eval-{fixture.name}"
+
+    try:
+        lf.create_dataset(
+            name=dataset_name,
+            description=fixture.description,
+            metadata={"type": "journey", "data_types": fixture.data_types},
+        )
+    except Exception:
+        pass  # Dataset may already exist
+
+    item_id = f"{fixture.name}--journey"
+    try:
+        lf.create_dataset_item(
+            dataset_name=dataset_name,
+            id=item_id,
+            input={
+                "directory": fixture.directory,
+                "data_types": fixture.data_types,
+                "user_messages": fixture.user_messages,
+            },
+            expected_output={
+                "criteria": fixture.expected_template_criteria,
+            },
+            metadata={"type": "journey"},
+        )
+    except Exception as exc:
+        logger.warning("langfuse_eval: failed to upsert journey dataset item %s: %s", item_id, exc)
+
+    lf.flush()
+    logger.info("langfuse_eval: synced journey fixture '%s' → dataset '%s'", fixture.name, dataset_name)
+    return dataset_name
+
+
 def create_eval_run(
     dataset_name: str,
     run_name: str,