James-Muthama/Carely-AI

Carely AI is a dual-agent CRM platform for African SMEs, automating WhatsApp customer support via RAG-powered Llama models and Groq inference. It turns uploaded PDFs into instant, accurate responses while a second agent classifies messages, scores sentiment, and flags knowledge gaps to drive continuous improvement. # Carely AI 🚀 **Carely AI** is an advanced, dual-agent Customer Relationship Management (CRM) platform. Originally conceptualized as an agentic-powered CRM solution for a hackathon, Carely bridges the gap between automated customer support and actionable business intelligence. By leveraging Retrieval-Augmented Generation (RAG) and high-speed LLM classification, Carely not only answers customer queries based on uploaded company documents but also continuously analyzes conversation logs to visually report on sentiments, categorize inquiries, and autonomously identify gaps in a business's knowledge base. --- ## ✨ Key Features ### 1. Customer Support Agent (RAG) - **Context-Aware Responses:** Automatically answers customer inquiries (via WhatsApp integration) using semantic search against uploaded business PDFs. - **Smart Fallbacks:** If a customer asks a question outside the scope of the uploaded documents, the agent gracefully handles it and tags the conversation for review. - **Persistent Memory:** Maintains conversation history for context-aware, multi-turn interactions. ### 2. Business Analytics Agent & Dashboard - **Real-Time KPI Tracking:** Tracks total conversations, message volume, overall customer sentiment, and top interaction categories. - **Dynamic Data Visualization:** Utilizes Chart.js for real-time line charts, doughnut charts, and bar charts. - **Category Management:** Tracks specific types of inquiries (e.g., *Pricing Inquiries, Technical Support, Complaints*). ### 3. AI-Driven Knowledge Gap Discovery - **Insight Generation:** Scans recent "Uncategorized" customer chats and cross-references them against the existing PDF knowledge base to identify missing information. - **Actionable Recommendations:** Automatically suggests new tracking categories and specific FAQ documents the business should create to improve the RAG agent's accuracy. - **Background Re-categorization:** When a new category is approved, a background thread seamlessly scans his …