FinAssure

Status	Rule Name	Threshold	Observed Value	Description
Run an experiment to populate release gate checklist.

Chaos Experiment Scenario Matrix

Scenarios mapped to Agentic Chaos fault primitives testing resilience boundaries.

Scenario ID	Fault Type	Targeted Step / Component	Hypothesis	Unsafe Outcomes

Test Runs

0

No runs found.

Select a run from the list

AgenticLens Span Waterfall

Click any run on the left to inspect its AgenticLens spans and Agentic Chaos event markers.

Version Comparison: v1 (Naive) vs v2 (Hardened)

Quantifying reliability gains against operational overhead (tokens, cost, latency).

Reliability & Overhead Delta Table

Dimension	v1 (Naive)	v2 (Hardened)	Delta / Impact

Disparate Resilience by Synthetic Segment

Verifying that fault resilience and silent-failure rates do not disproportionately affect specific applicant groups.

Synthetic Group	Total Applications	Silent Failure Rate	Safe Degradation Rate	Escalation Rate

Decision & Adverse-Action Audit

Inspect individual applicant decisions, verify deterministic oracle ground truth vs model determinations, and validate adverse notice grounds.

Select Case:

📋 Application Inputs

Select a case above to inspect input details.

Financial Documents Preview:

No document selected

⚖️ Deterministic Policy Oracle

Pure-Python ground truth evaluation. No LLM arithmetic.

🤖 System Determination

Multi-agent pipeline outcome, factors, narrative, and escalation audit.

Multi-Agent Workflow Topology

Interactive topology showing agent handoffs, tool invocation edges, and injected chaos fault points.

TopologyTracker Active

PIPELINE EXECUTION FLOW & FAULT INTERCEPTORS

Agent Span Tool Call Human Review Chaos Fault Point

Click on any agent node or fault marker above to inspect its contract and fault behavior.

Resource, Cost & Latency Profiling

Measured via AgenticLens profiler per agent span: token volume, API pricing, and latency SLAs.

Cost Per Application

$0.0028

SLA Budget: ≤ $0.0500 / app

p95 Latency Performance

180 ms

SLA SLA Threshold: ≤ 4000 ms

Tokens Per Application

~280 tokens

Deterministic arithmetic offloaded from LLM

Token Usage By Agent Span

Agent Name	Agent Role	Type	Tokens	Avg Latency	Cost Contribution

Interactive Scenario Launcher

Select chaos scenarios, tune fault injection parameters, and execute the reliability test harness.

Target Workflow Pack

Pipeline Version

Synthetic Applicants Count Full matrix support up to 200 cases

Reps Per Case

Select Chaos Scenarios To Execute

Live Execution Output

Ready to execute. Select scenarios and click "Execute Reliability Run".

Audit-Ready Evidence Packages

Reports generated per release mapped to DORA scenario testing and AI Act human-oversight principles.

Framework Mapping Context:

• DORA (Digital Operational Resilience Act): Validates graceful escalation when downstream dependency services (Credit Bureau / Sanctions watchlists) fail.

• EU AI Act Human Oversight: Enforces refer-to-human review triggers and caps automated decision-making on incomplete or degraded evidence.

• Adverse Action Explainability: Adverse-action reasons are computed through deterministic arithmetic rules to eliminate hallucinated rejections.

Release Gate Status

Automated Release Gate Policy Rules