/ Build log / Live ledger
Watch the ledger.
Most AI projects have logs. VoidAi has an append-only ledger. Every action any of the 19 agents takes writes one row. Cost. Latency. Outcome. Run ID. Agent. One JSONL line per action.
Below is a sanitized snapshot of the most recent activity. Run IDs are truncated to their first 8 chars. File targets and paths are redacted. The point is to make the governance layer observable -- not to leak operational state.
Snapshot taken
no snapshot yet
Source
state/audit.jsonl (last 0 rows)
Refresh cadence
on every site deploy
Rows shown
0
Recent activity
Snapshot pending. The
public ledger snapshot hasn't been generated for this deploy
yet. Each deploy regenerates a sanitized copy of the audit log
to public/data/audit-snapshot.json;
on the next deploy this table fills in.
To generate the snapshot manually before deploying:
.\scripts\build-audit-snapshot.ps1
Then re-run npm run deploy from C:\AI-Company\web.
Schema
Each row is a JSON object on one line. Required keys:
{
"ts": "2026-04-25T20:09:24.604Z", // UTC ISO 8601
"run_id": "<uuid>",
"agent": "void-pulse",
"action": "pulse.probe",
"target": "env", // can be a path
"outcome": "success", // success|failure|allowed|denied|partial|skipped
"cost_usd": 0.0001, // float
"duration_ms": 80,
"reason": "status=green;detail=...", // semicolon-delimited k=v pairs
"playbook": "pulse-probe",
"playbook_version": "0.1.0"
}
Sanitization rules for this public mirror: run_id truncated to first 8 chars, target redacted to its file extension or "(redacted)", reason truncated to 80 chars, inputs_hash dropped entirely. Source rows preserved verbatim in the private audit log.
If your agents need this layer
Most AI consultancies ship the agents. Few ship the ledger. We install the ledger underneath your existing stack. Audit log, capability gates, daily caps, shadow-window promotions.