VFS Tactical Operations Manual

Version: 1.5 | Status: Effective | Updated: 2026-01-03

Standard Operating Procedures (SOPs) for the steering and maintenance of the VFS Intelligence Engine.

1. Rapid Ingestion & Triage

SOP-INGEST-01

Procedure for forcing manual ingestion of a high-priority topic.

Trigger: Click the Ingest (Plus) icon in the header.
Verification: Monitor the Deployment Monitor tab for the new topic transitioning from triage to analyzed.
Action: If a topic hangs, check the Cloud Run logs for vfs-triage timeouts.

SOP-PUBLIC-05

Verification protocol for public-facing event cards, evidence trays, and entity profiles.

Audit Flow: Navigate to the Public UI and locate the topic card.
Grounding Check: Expand 🛡️ GROUNDED CLAIMS and click any Cyan Entity Link.
Entity Profile: Verify the Adjudication Ledger contains the current topic and a calculated Consistency Score.
Show Work: Click 👁️ SHOW WORK to verify the recursive neural trace and subject resolution confidence.

SOP-AUTO-02

Ensuring global credibility scores are calculated and synchronized.

Interval: The vfs-analytics-write-back Cloud Scheduler job runs every 4 hours.
Manual Trigger: In the Credibility Matrix tab, click Trigger Write-Back to force a BigQuery-to-Firestore synchronization.
Cost Check: Monitor Resource Oversight for "Automation Overhead" costs ($0.0001 per run).

SOP-NORM-07

Promoting discovered candidates to authorized headings and unifying historical data.

Promotion: In the Credibility Matrix, click Promote on a "Candidate" entity. Search and link to the canonical Library of Congress (LCSH) record.
Refinement: Add known spelling variants (e.g., "Trump" or "DJT") to the authority record to capture future mentions.
Normalization: Click Normalize History to anchor all historical claims to the current authority map via the dashboard.
Cleanup: Click Trigger Write-Back. The analytics service will automatically purge any stale "ghost" rows from the matrix by synchronizing with current BigQuery results.

SOP-EMERGENCY-99

Immediate mitigation of toxic or hallucinated content.

Block Source: In the Source Guardrails view, update the status of a problematic domain to Blocked.
Cleanup: If a topic is invalid, click Reject in the Monitor view to archive the proposal and prevent public propagation.