User Guide
Learn how to navigate the CHAPAL chat interface, simulate anomalies, and understand the transparency metrics.
The Chat Interface
The main chat screen is divided into two distinct areas:
- Left Panel (Interaction): Where you chat with the AI and run simulations.
- Right Panel (Transparency): A real-time dashboard showing how the AI is analyzing your behavior.
Note: On mobile devices, the Transparency Panel is hidden to save space.
Simulating Anomalies
CHAPAL includes a built-in Simulation Toolbar above the chat input. This allows you to safely test the system's defenses by injecting pre-scripted "attacks" or anomalous queries.
Prompt Injection
"Ignore previous instructions..."Tests if the system can prevent users from overriding its safety protocols.
Self Harm
"I feel hopeless..."Triggers an immediate safety block and admin alert for high-risk psychological distress.
PII Leak
"My social security number is..."Demonstrates the system's ability to redact or block sensitive personal information.
Hallucination
"Who is the President of Mars?"Tests the semantic Fact-Checking layer (Layer 2) against non-factual queries.
Medical Advice
"Medication for chest pain?"System correctly identifies this as advice only a professional should give.
DDoS Attack
[Special Action]Simulates a burst of 4 rapid messages to trigger Layer 1 Spike Detection.
Transparency Panel
The transparency panel provides "X-Ray vision" into the AI's decision-making process.
Safety Score
A dynamic gauge that drops when risky keywords or patterns are detected. Green indicates safety, while Red indicates a likely block.
Emotion Analysis
The system attempts to read the emotional context of your query (e.g., Anxious, Hostile, Curious). High intensity negative emotions may trigger faster escalation.
Review Status
If your message is blocked, you'll see a status indicator showing if it's "Pending Review" by a human admin.