Evidence Viewer

The Evidence Viewer lets you deep-dive into any governance decision. It renders the full TEEC (Typed Evidence & Evidence Contract) v2.1 envelope, showing exactly why a decision was made, which modules evaluated, how confident the detections were, and the cryptographic proof of decision integrity.

What It Shows

For any governance decision, the Evidence Viewer displays:

Section	Contents
Top Fields	correlationId, agentId, timestamp, teec_version, action, intent_ref, receipt_ref, seq, running_count
Reason Chain	Ordered module evaluations grouped by stage (pre/post), with per-module action, latency, and reason codes
Confidence Scores	Per-finding confidence values (0–100%) with color-coded severity indicators
Governance Seal	HMAC seal, agent_id, timestamps, receipt chain, genesis detection
Cost Evidence	Provider, model, token breakdown, estimated vs actual cost, variance warnings

Navigating to the Evidence Viewer

You can reach the Evidence Viewer from three entry points: Navigate to /evidence to see the filterable decision list. Click any decision to open its evidence detail.

2. From Audit Trail

Click any governance event in the Audit Trail → opens directly to the Evidence detail for that correlationId.

3. From Decision Explorer

Click “View Full Evidence” in any Decision Explorer drill-down → opens the Evidence detail.

Reason Chain Visualization

The reason chain shows the ordered sequence of module evaluations:

┌─────────────────────────────────────────────────────┐
│ PRE_EXECUTION                                       │
├─────────────────────────────────────────────────────┤
│ 🟢 TealSecrets v1.2.0    PASS   2.1ms  (no findings)│
│ 🔴 TealGuard v1.4.0      DENY   3.4ms              │
│    └── PII_DETECTED (high) — SSN pattern found      │
│    └── INJECTION_SUSPECTED (medium)                  │
├─────────────────────────────────────────────────────┤
│ POST_EXECUTION                                       │
├─────────────────────────────────────────────────────┤
│ (not evaluated — pre-scan denied)                    │
└─────────────────────────────────────────────────────┘

Final Action: DENY (most restrictive: TealGuard)

Each module evaluation shows:

Module name and version
Stage (PRE_EXECUTION or POST_EXECUTION)
Action taken by this module
Latency contribution
Reason codes with title and severity from the TEEC registry
Error indicator (if the module errored)

Evaluations are grouped into PRE_EXECUTION and POST_EXECUTION sections with visual separation.

Confidence Scores

When modules produce findings with confidence scores, they’re displayed as percentage bars:

Finding	Confidence	Level
SSN Pattern	94%	🔴 High (> 70%)
Email Address	99%	🔴 High (> 70%)
Possible Phone	52%	🟡 Medium (40–70%)
Ambiguous Token	28%	⚪ Low (< 40%)

Classification:

High (> 70%): Strong signal — likely a true positive
Medium (40–70%): Moderate signal — review recommended
Low (< 40%): Weak signal — possible false positive

Findings are sorted by confidence in descending order (highest confidence first).

Governance Seal

The seal section proves tamper-evidence:

Field	Example	Description
HMAC	`a7f3...b2c1` (truncated)	HMAC-SHA256 over decision payload
Seal Timestamp	2026-06-15 14:32:01.234	When the seal was computed
Agent ID	`research-agent-01`	Agent that produced the decision
Time Diff	+12ms	Difference between seal and decision timestamp
intent_ref	`sha256:abc...`	Hash linking to the original intent
receipt_ref	`sha256:def...`	Hash linking to the previous decision
seq	47	Position in the decision chain
normalization_id	`norm-001`	Normalization identifier

Chain Genesis Detection

When receipt_ref is exactly 64 zeros (0000...0000), the viewer displays a “Chain Genesis” badge indicating this is the first decision in the agent’s chain.

Copy Seal

The “Copy Seal” button copies the full hex-encoded HMAC value to the clipboard for verification.

Cost Evidence

When cost data exists for the decision:

Field	Value
Provider	OpenAI
Model	gpt-4o
Input Cost	$0.0025
Output Cost	$0.0040
Total Cost	$0.0065
Request Tokens	500
Response Tokens	800
Total Tokens	1300

Estimated vs Actual

When both values are available:

Estimated: $0.0050
Actual:    $0.0065
Variance:  +$0.0015 ⚠️ (+30% over estimate)

A warning indicator appears when:

Actual cost exceeds estimated by more than 20%
Estimated cost is zero or negative

No Cost Data

If no cost record exists for the correlationId, the section displays: “No cost data available.”

Filtering

The Evidence list view includes a filter panel with:

Filter	Type	Options
Reason Codes	Multi-select	Populated from distinct reason codes in the data
Module Name	Dropdown	Populated from distinct module names in the data
Action Type	Dropdown	ALLOW, DENY, SANITIZE, REPORT
Agent ID	Text input	Partial match (contains)

Filter Behavior

All active filters use AND logic — results must match all active filters simultaneously
Active filter count and result count are displayed
Clear Filters button resets all filters at once

Export

Export Evidence Pack

Every evidence detail view includes an “Export Evidence Pack” button that downloads the complete TEEC envelope as JSON: Filename: teec-evidence-{correlationId}.json Contents:

{
  "metadata": {
    "exportTimestamp": 1718461921000,
    "dashboardVersion": "1.4.0",
    "teec_version": "2.1"
  },
  "decision": {
    "correlationId": "abc-123-def",
    "agentId": "research-agent-01",
    "timestamp": 1718461920000,
    "action": "DENY",
    "intent_ref": "sha256:...",
    "receipt_ref": "sha256:...",
    "seq": 47,
    "running_count": 48
  },
  "moduleEvaluations": [...],
  "governanceSeal": {
    "hmac": "a7f3...full hex...",
    "timestamp": 1718461920012,
    "agent_id": "research-agent-01"
  },
  "costEvidence": {
    "provider": "openai",
    "model": "gpt-4o",
    "totalCostUsd": 0.0065,
    ...
  }
}

This file serves as a tamper-proof audit artifact for compliance teams.

Error Handling

If the export fails (network issue, browser restriction), a toast notification appears with:

Error message explaining what went wrong
Retry button to attempt the download again

API Endpoints

The Evidence Viewer is powered by:

Endpoint	Purpose
`GET /api/v1/decisions/{correlationId}/evidence`	Assemble and return the full TEEC envelope
`GET /api/v1/decisions/{correlationId}/evidence/export`	Download the envelope as a JSON file
`GET /api/v1/decisions?reasonCode=&moduleName=&actionType=&agentId=`	Filtered decision list

Partial Data

The API returns HTTP 200 with partial data when some components are missing:

Decision exists but no module evaluations → moduleEvaluations: []
Decision exists but no cost record → costEvidence: null
Governance seal unparseable → governanceSeal: null

The Evidence Viewer gracefully handles missing sections, showing “No data available” messages where appropriate.

Getting Started

Core Concepts

Governance Domains

Policy Authoring

Architecture

Integrations

Deployment

API Reference

Cookbook

Compliance & GRC

Guides

Reference

Versions

Playground

Dashboard

Blog

Evidence Viewer

Evidence Viewer

What It Shows

Navigating to the Evidence Viewer

1. Direct Navigation

2. From Audit Trail

3. From Decision Explorer

Reason Chain Visualization

Confidence Scores

Governance Seal

Chain Genesis Detection

Copy Seal

Cost Evidence

Estimated vs Actual

No Cost Data

Filtering

Filter Behavior

Export

Export Evidence Pack

Error Handling

API Endpoints

Partial Data

​Evidence Viewer

​What It Shows

​Navigating to the Evidence Viewer

​1. Direct Navigation

​2. From Audit Trail

​3. From Decision Explorer

​Reason Chain Visualization

​Confidence Scores

​Governance Seal

​Chain Genesis Detection

​Copy Seal

​Cost Evidence

​Estimated vs Actual

​No Cost Data

​Filtering

​Filter Behavior

​Export

​Export Evidence Pack

​Error Handling

​API Endpoints

​Partial Data

Evidence Viewer

What It Shows

Navigating to the Evidence Viewer

1. Direct Navigation

2. From Audit Trail

3. From Decision Explorer

Reason Chain Visualization

Confidence Scores

Governance Seal

Chain Genesis Detection

Copy Seal

Cost Evidence

Estimated vs Actual

No Cost Data

Filtering

Filter Behavior

Export

Export Evidence Pack

Error Handling

API Endpoints

Partial Data