Epic 6.2: SLOs - Error Budgets, Burn Rates, Multi-Window Alerting (#51)

* feat: Epic 6.2 - SLO module with error budgets, burn rates, and MCP tool

- SLOTarget dataclass with validation, ErrorBudget with float-tolerant exhaustion
- BurnRateCalculator with Google SRE multi-window alerting model
- SLOEvaluator reads from MetricsRegistry via Prometheus .collect() API
- 3 default targets: latency_p99 <30s, error_rate <5%, availability >99%
- check_slos MCP tool (8 total tools)
- 35+ new tests (1,858 total passed)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* fix: address all review findings - 5 agents, 12 fixes, 22 new tests

Security:
- Wrap check_slos in try/except + handle_mcp_exception (error oracle)
- Clamp burn_rate to MAX_BURN_RATE=1000.0 (no float('inf') in JSON)
- Add SLOTarget validation: non-empty name, finite non-negative threshold
- Filter _get_request_counts to _total suffix (PROMETHEUS_ENABLE_CREATED_SERIES safety)

Correctness:
- Fix objective=1.0 contradictory state (exhausted + 100% remaining)
- Fix p99 multi-label aggregation (sum by le across agent × status)
- Add p99 linear interpolation between bucket boundaries
- Fix p99 overflow: return highest bucket boundary instead of None
- Correct docstring: simplified burn-rate ratio, not Google multi-window

Hardening:
- DEFAULT_TARGETS immutable tuple (was mutable list)
- SLOTarget frozen=True dataclass
- ErrorBudget zero-budget path consistent status

Tests: 22 new adversarial tests (1,884 total passed)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* fix: R2 hardening - NaN-safe JSON, isfinite guard, coupling docs

- json.dumps(allow_nan=False) prevents NaN/Infinity in JSON output
- math.isfinite() guard on p99 interpolation result
- ValueError/OverflowError catch in _get_request_counts
- Cross-reference comment for status label coupling (execution.py)

R2 review: 2 APPROVE (sec, opus), 2 REQUEST_CHANGES on documented
design limitations (in-process counters, hardcoded dispatch) - v2 items.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Brian Krafft <bkrafft@microsoft.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Author: 3 people

RUNBOOK.yaml

@@ -1,8 +1,8 @@
 schema_version: '1.0'
 lock:
-  active_claim: null
-  claimed_by: null
-  claimed_at: null
+  active_claim: '6.2'
+  claimed_by: 2026-04-07-e9b4572b
+  claimed_at: '2026-04-07T00:29:00Z'
 review:
   max_rounds: 3
   required_gates:
@@ -547,8 +547,8 @@ epics:
 - id: '6.2'
   phase: P6
   title: SLOs (latency, error rate, availability)
-  status: pending
-  branch: null
+  status: claimed
+  branch: epic/6.2-slos
   pr: null
   review_round: 0
   description: 'Define and enforce SLO targets: p99 execution latency, error rate

openspace/mcp/tool_handlers.py

@@ -829,6 +829,25 @@ async def get_execution_traces(limit: int = 5) -> str:
     return json.dumps([t.to_dict() for t in traces], indent=2, default=str)
 
 
+async def check_slos() -> str:
+    """Check current SLO (Service Level Objective) compliance.
+
+    Returns current status for all SLO targets including error budget
+    remaining, burn rate, and any active alerts.  Use this to monitor
+    system health against defined reliability targets.
+    """
+    try:
+        from openspace.observability.metrics import metrics
+        from openspace.observability.slos import SLOEvaluator
+
+        evaluator = SLOEvaluator(registry=metrics)
+        return evaluator.to_json()
+    except Exception as e:
+        from openspace.errors import EXECUTION_ERROR, handle_mcp_exception
+
+        return handle_mcp_exception(e, tool_name="check_slos", error_code=EXECUTION_ERROR)
+
+
 # ---------------------------------------------------------------------------
 # Registration
 # ---------------------------------------------------------------------------
@@ -845,3 +864,5 @@ def register_handlers(mcp) -> None:
     mcp.tool()(health_check)
     mcp.tool()(get_metrics)
     mcp.tool()(get_execution_traces)
+    # SLOs (Epic 6.2)
+    mcp.tool()(check_slos)

openspace/observability/__init__.py

@@ -1,13 +1,15 @@
-"""openspace.observability — Metrics, tracing, and health for OpenSpace.
+"""openspace.observability — Metrics, tracing, health, and SLOs for OpenSpace.
 
 Provides Prometheus-compatible metrics, structured execution traces,
-and aggregated health checks for the MCP server and grounding engine.
+aggregated health checks, and SLO tracking for the MCP server and
+grounding engine.
 
-Epic 6.1 — Phase 6 (Operability).
+Epics 6.1–6.2 — Phase 6 (Operability).
 """
 
 from openspace.observability.health import HealthAggregator, HealthStatus
 from openspace.observability.metrics import MetricsRegistry
+from openspace.observability.slos import SLOEvaluator
 from openspace.observability.tracing import ExecutionTracer, trace_async
 
 __all__ = [
@@ -16,4 +18,5 @@
     "trace_async",
     "HealthAggregator",
     "HealthStatus",
+    "SLOEvaluator",
 ]