Skip to content

AI Assistant

CalcBridge includes an AI-powered chat assistant for natural language interaction with portfolio data.

Overview

The AI assistant allows users to:

  • Ask questions about workbook data in plain English
  • Get contextual answers based on uploaded portfolio information
  • Receive financial analysis and metric calculations on demand

How It Works

flowchart LR
    USER["User Query"] --> SCRUB["PII\nScrubbing"]
    SCRUB --> AI["SecureAI\nGateway"]
    AI --> SANITIZE["Response\nSanitization"]
    SANITIZE --> RESPONSE["User\nResponse"]

    style SCRUB fill:#FEF3C7,stroke:#F59E0B
    style AI fill:#DBEAFE,stroke:#3B82F6

Request Flow

  1. User submits query via the chat endpoint
  2. PII scrubbing removes sensitive data before AI processing
  3. SecureAIGateway processes the query with workbook context
  4. Response sanitization ensures no sensitive data leaks
  5. Sanitized response returned to the user

Usage

Basic Chat

POST /api/v1/ai/chat
{
  "query": "What is the weighted average spread of my portfolio?",
  "context": {
    "workbook_id": "550e8400-e29b-41d4-a716-446655440000"
  }
}

Context-Aware Queries

The assistant understands CLO and financial terminology:

  • "What percentage of the portfolio is CCC-rated?"
  • "Which loans mature in the next 6 months?"
  • "Show me the top 5 obligors by par exposure"
  • "What is the current OC test cushion?"

PII Protection

Before any data is sent to the AI model:

  • Borrower names are anonymized
  • Account numbers are masked
  • Email addresses are removed
  • Custom PII patterns can be configured per tenant

Limitations

Limitation Detail
Query length Maximum 1000 characters
Response scope Based on currently uploaded workbook data
Real-time data Does not access live market data
Advice Informational only, not financial advice

Security

  • All AI interactions are tenant-isolated
  • Chat history is per-user
  • No data is stored by external AI providers
  • Responses are sanitized to prevent information leakage