Audit & Metrics
Prove what was allowed, masked, or blocked on act paths. Metadata-first audit log records governance outcomes per event. Execution metrics and compileHash tie those events back to runs and policy.
This page is the proof chapter after Run in production. Use the audit log for investigations. Use execution metrics for trends. Every governed mask and enforce record carries compileHash so you can tie events back to published runtime policy.
What gets recorded
Mask, enforce, and client ingress paths emit metadata-only records. Responses on governed APIs include compileHash when policy was evaluated.
| Field | Meaning |
|---|---|
compileHash | Fingerprint of the published policy bundle evaluated for this event. Matches run rows and policy stream. |
verdict / action | Governance outcome: allow, mask, block, audit (observe), or hold when review is required. |
surface | Where enforcement ran: local hooks, hosted ingress (Platform Team), sidecar, or cluster cells. |
clientSurface | Caller identity (SDK, MCP, IDE hook, automation) for routing and scope audits. |
run / workflow | Optional act identifiers when the event belongs to a cell-backed workflow run. |
timestamp | UTC time the control plane accepted the record (retention follows your plan and contract). |
Audit log in the app
Open Monitor → Audit log for the searchable event stream. Each row should be enough to answer who called, which surface enforced, and what verdict applied, without opening raw customer content in the console.
Pair audit with Runs when you need act status, cell placement, and duration. Runs store compileHash at birth.
Filters and investigations
Investigators usually move from a symptom (unexpected block, missing mask) to a narrow audit slice. Use filters in the app before exporting or forwarding metadata.
| Filter | Use when |
|---|---|
| compileHash | Prove which published policy governed an act run or enforce path. |
| verdict | Find masks, blocks, or holds during an incident window. |
| surface | Separate SDK/MCP ingress from cluster cells when debugging placement. |
| time range | Narrow audit to a release window or customer-reported timestamp. |
| workflow / run id | Drill from a single Runs row into related audit lines. |
Typical drill-down:
- Open the run in Runs and copy
compileHash. - Filter audit log by that hash and compare with published runtime policy or the policy stream API.
- If hash matches but verdict surprises you, simulate with SDK
simulatePolicy()before changing production guards. - If hash differs from expectation, check whether a publish or approval is still pending in Threat signals & approvals.
Execution metrics
Monitor → Execution metrics summarizes throughput and policy outcomes. Use it for weekly reviews and capacity conversations, not as the only source during a single-run incident.
| Console view | Purpose | Not a substitute for |
|---|---|---|
| Audit log | Event-level proof with compileHash, verdict, and surface per row. | Capacity planning alone (pair with execution metrics). |
| Execution metrics | Aggregated throughput and policy outcomes over time. | Per-incident drill-down (open audit log with filters). |
| Workspace usage | Included act run-minutes and guarded-token meters for billing. | Enforcement proof (see Runs and audit log). |
Runs and compileHash
The same compileHash appears on run rows, audit events, and policy fetch responses. When all three align, you have a closed loop: the act you started used the policy you think is live. After publish, confirm new runs pick up the new hash before you declare the change safe.
Authoring and staging live in Runtime policy. Workspace token checks for automation are in API keys.
Export and forwarding
Push metadata from automation with POST /api/events (events:write scope) or the SDK emitEvent() helper. See Events API and Telemetry and events for field shapes aimed at SIEM and GRC pipelines.
import { BleklineClient } from "@blekline/client";
const client = new BleklineClient();
await client.emitEvent({
kind: "governance.client",
platform: "ci",
action: "audit",
clientSurface: "sdk",
entitiesMasked: 0,
});Enterprise workspaces can use Compliance export for batch metadata extracts. Connector-specific SIEM setup belongs in your runbook; describe trust boundaries in Security before procurement reviews.