feat(F37-TRACE-003): data model integration for traces layer

- Add trace.schema.json: Traces layer schema (id, sprint_id, lm_calls array, cost/latency, status)
- Add model/traces.json: Initial traces layer (empty)
- Register traces_router in api/routers/layers.py
- Add traces imports/registration in api/server.py
- GET  /model/traces/ - list all traces (paginated)
- GET  /model/traces/{correlation_id} - retrieve single trace
- PUT  /model/traces/{correlation_id} - upsert trace (auto row versioning, audit fields)
- DELETE /model/traces/{correlation_id} - soft delete
- Verified endpoints with test trace: PUT + GET + LIST all OK (200 responses)
- Cosmos persistence enabled (ACA production) + MemoryStore (local dev)
- Ready for lm_tracer output persistence (Phase 3 complete)

Author: Presta

api/routers/layers.py

@@ -41,3 +41,4 @@
 milestones_router      = make_layer_router("milestones",       "/model/milestones",       ["milestones"])
 risks_router           = make_layer_router("risks",            "/model/risks",            ["risks"])
 decisions_router       = make_layer_router("decisions",        "/model/decisions",        ["decisions"])
+traces_router          = make_layer_router("traces",           "/model/traces",           ["traces"])

api/server.py

@@ -252,6 +252,7 @@ async def dispatch(self, request: Request, call_next):
         # project plane (E-07/E-08)
         projects_router, wbs_router,
         sprints_router, milestones_router, risks_router, decisions_router,
+        traces_router,
     )
     from api.routers.fp import router as fp_router  # noqa: E402
     from api.routers.filter_endpoints import router as endpoints_router
@@ -275,6 +276,7 @@ async def dispatch(self, request: Request, call_next):
         # project plane
         projects_router, wbs_router,
         sprints_router, milestones_router, risks_router, decisions_router,
+        traces_router,
         fp_router,
         # graph (E-11)
         graph_router,

model/traces.json

@@ -0,0 +1,4 @@
+{
+  "$schema": "../schema/trace.schema.json",
+  "traces": []
+}

schema/trace.schema.json

@@ -0,0 +1,86 @@
+{
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "$id": "trace.schema.json",
+  "title": "EVA LM Trace",
+  "description": "End-to-end trace of LM calls from a single correlation_id. Captures model, tokens, cost, and latency per call. Partition key for Cosmos: correlation_id.",
+  "type": "object",
+  "required": ["id", "sprint_id", "created_at", "status"],
+  "additionalProperties": false,
+  "properties": {
+    "id": {
+      "type": "string",
+      "description": "Correlation ID, format: SPRINT-{sprint_id}-{timestamp}-{uuid[:8]}",
+      "examples": ["SPRINT-0.5-20260301143021-a1b2c3d4"]
+    },
+    "sprint_id": {
+      "type": "string",
+      "description": "Reference to sprint.id from sprints layer"
+    },
+    "story_id": {
+      "type": ["string", "null"],
+      "description": "Story/bug ID being traced, e.g. BUG-F37-001"
+    },
+    "phase": {
+      "type": ["string", "null"],
+      "enum": ["A", "B", "C", null],
+      "description": "Bug fix phase: A=RCA, B=Code, C=Prevention. Null if multi-phase aggregate."
+    },
+    "created_at": {
+      "type": "string",
+      "format": "date-time",
+      "description": "When trace was created (RFC3339)"
+    },
+    "lm_calls": {
+      "type": "array",
+      "description": "Array of LM interactions",
+      "items": {
+        "type": "object",
+        "required": ["model"],
+        "properties": {
+          "model": {
+            "type": "string",
+            "enum": ["gpt-4o", "gpt-4o-mini", "claude-3-opus", "claude-3-haiku"],
+            "description": "Model called"
+          },
+          "tokens_in": {
+            "type": ["integer", "null"],
+            "description": "Input tokens consumed"
+          },
+          "tokens_out": {
+            "type": ["integer", "null"],
+            "description": "Output tokens generated"
+          },
+          "cost_usd": {
+            "type": ["number", "null"],
+            "description": "Cost in USD (8 decimal places)"
+          },
+          "latency_ms": {
+            "type": ["integer", "null"],
+            "description": "API call latency in milliseconds"
+          }
+        }
+      }
+    },
+    "total_cost_usd": {
+      "type": ["number", "null"],
+      "description": "Sum of all LM call costs (USD, 8 decimals)"
+    },
+    "total_latency_ms": {
+      "type": ["integer", "null"],
+      "description": "Sum of all LM call latencies (ms)"
+    },
+    "status": {
+      "type": "string",
+      "enum": ["active", "archived", "failed"],
+      "description": "Trace lifecycle: active (current), archived (>90d), failed (error during collection)"
+    },
+    "is_active": {
+      "type": "boolean",
+      "description": "Whether this trace is queryable (false after archival)"
+    },
+    "notes": {
+      "type": ["string", "null"],
+      "description": "Operator notes or error messages"
+    }
+  }
+}