FinCaseAI is an AI-powered credit dispute assistant that uses Perplexity’s Sonar Reasoning API to analyze financial dispute cases and suggest approve/reject decisions with detailed reasoning. It helps analysts by providing fraud detection, sentiment scoring, and urgency assessment for credit card dispute complaints.

​
Features

  • Approve/Reject Decision Suggestions with confidence levels and detailed reasoning
  • Fraud Risk Scoring using keyword and pattern matching algorithms
  • Sentiment and Urgency Detection from complaint text analysis
  • Sonar API Integration for intelligent recommendations and structured reasoning
  • Optional Redis Caching for improved performance and response times
  • Structured Output with citations and supporting evidence

​
Prerequisites

FinCaseAI sends case descriptions as structured prompts to the Sonar API, asking for:
  • A decision recommendation (approve/reject)
  • A brief explanation for the judgment
The API response is parsed and shown alongside internal fraud and sentiment scores for a complete picture.
Example prompt to Sonar 
Given the following customer complaint, should the dispute be approved or rejected? Provide a short explanation.

​
Tech Stack

  • Frontend: React + Vite
  • Backend: Python Flask
  • AI Engine: Perplexity Sonar API + custom logic
  • Cache: Optional Redis integration
  • Infrastructure: Runs locally or deployable via Docker

​
Setup Instructions

  1. Clone the GitHub repo
  2. Create a .env file with your PERPLEXITY_API_KEY
  3. Run the backend using Flask
  4. Start the frontend using Vite
  5. Open http://localhost:5173 and start testing your cases