VerixAI: Intelligent Document Analysis with Citations

VerixAI is a production-ready document analysis platform focused on one core mission: helping knowledge workers query large document collections efficiently and receive AI-generated answers with precise, verifiable citations.

Designed for professionals handling extensive document repositories—whether medical records, legal cases, or corporate policies—VerixAI provides accurate, context-aware responses backed by advanced retrieval and ranking algorithms.

---

✨ Key Features

• Multi-Format Document Ingestion: Process PDF, DOCX, PPTX, HTML, TXT, MD, XLSX, and JSON files with high-fidelity extraction.
• Advanced RAG Pipeline: Sophisticated Retrieval-Augmented Generation with hybrid search (semantic + keyword) and advanced ranking algorithms.
• Precise Citations: Every answer includes clear citations with source documents, chunk indices, and confidence scores for full verifiability.
• Role-Aware Responses: Tailored responses for different professional contexts (Doctor, Lawyer, HR) with appropriate disclaimers and terminology.
• Dataset Management: Organize documents into isolated, searchable collections for improved query accuracy and context control.
• Multi-Agent Architecture: Specialized AI agents for document ingestion, parallel retrieval, ranking, citation validation, and quality control.
• Parallel Processing: Asynchronous document processing with configurable worker pools for handling large-scale document collections.
• Scalable Infrastructure: Modern async stack (FastAPI, React) containerized for easy deployment and horizontal scaling.

---

🚀 Live Demo & Screenshots

✨ Check out the live demo here! ✨

Query Interface

Upload and Dataset Management

---

🗺️ Roadmap

Recently Completed ✅

• Multi-Agent Architecture: Specialized agents for document processing and retrieval
• Advanced Retrieval System: Hybrid search with semantic and keyword matching
• Parallel Document Processing: Asynchronous processing with worker pools
• Multi-LLM Support: Support for Ollama (local models), OpenAI, and Claude
• Citation Validation: Automated validation of source citations

Phase 1: Core Enhancements (Q1 2025)

• Advanced Document Processing: OCR support for scanned documents
• Real-time Collaboration: Multiple users working on same dataset
• Export Functionality: Export Q&A sessions as reports (PDF/Word)
• Voice Input/Output: Speech-to-text queries and text-to-speech responses

Phase 2: Intelligence Features (Q2 2025)

• Document Comparison: Compare and contrast multiple documents
• Knowledge Graph: Visual representation of document relationships
• Custom Embeddings: Fine-tune embeddings for specific domains
• Advanced Analytics Dashboard: Usage metrics and insights

Phase 3: Enterprise Features (Q3 2025)

• SSO Integration: SAML/OAuth support for enterprise authentication
• Audit Logging: Complete audit trail for compliance
• API Rate Limiting: Advanced rate limiting and usage analytics
• Multi-tenancy: Support for multiple isolated organizations

Phase 4: Advanced Analytics (Q4 2025)

• Sentiment Analysis: Analyze document sentiment and tone
• Entity Recognition: Extract and link named entities
• Time-series Analysis: Track changes across document versions
• Custom Models: Support for domain-specific fine-tuned models

Community Contributions Welcome!

We encourage contributions in these areas:

• Additional file format support (epub, rtf, etc.)
• Language translations and internationalization
• Performance optimizations
• Bug fixes and documentation improvements

---

🛠️ Tech Stack & Architecture

VerixAI is built with a modern, microservices-oriented architecture.

Component	Technology
Backend	FastAPI, Python 3.10+, LangChain, Uvicorn
Frontend	React 18, TypeScript, Material-UI (MUI), Axios
Vector Database	ChromaDB
LLM & Embeddings	OpenAI (GPT-4, text-embedding-3-small)
Document Processing	MarkItDown
Infrastructure	Docker, Docker Compose, Nginx
Optional	PostgreSQL (for metadata)

System Architecture

┌─────────────────┐     ┌─────────────────┐
│                 │     │                 │
│  React Frontend │────▶│  Nginx Proxy    │
│ (Port 3000)     │     │ (Production)    │
│                 │     │                 │
└─────────────────┘     └────────┬────────┘
                                 │
                                 ▼
                    ┌────────────────────┐
                    │                    │
                    │  FastAPI Backend   │
                    │   (Port 8000)      │
                    └──┬──────┬──────┬──┘
                       │             │
                ┌──────▼──┐        ┌─▼────────┐
                │ChromaDB │        │PostgreSQL│
                └─────────┘        └──────────┘

---

🏁 Getting Started

The easiest way to get VerixAI running is with Docker and Docker Compose.

Prerequisites

• Docker and Docker Compose installed
• An OpenAI API Key
• Git

Installation

1. Clone the repository:

   git clone https://github.com/arunsai63/verix-ai.git
   cd verix-ai

2. Set up your environment variables:

   • Copy the example environment file:

     cp .env.example .env

   • Edit the .env file and add your OpenAI API key:

     OPENAI_API_KEY=your-actual-api-key-here
     # Generate a secure random key for JWTs
     SECRET_KEY=a_very_secure_random_string_of_at_least_32_characters
     

3. Launch the application with Docker Compose:

   docker-compose up --build -d

   This command builds the images and starts the frontend, backend, and ChromaDB services in detached mode.

4. Access VerixAI:

   • Frontend Application: http://localhost:3000
   • Backend API Docs: http://localhost:8000/docs

---

📖 Usage

1. Upload Documents: Navigate to the "Upload" tab, select or drag-and-drop your files, assign them to a new or existing dataset, and click "Upload".
2. Query Documents: Go to the "Query" tab, type your question, select the dataset(s) to search, choose a professional role, and get your cited answer.
3. Manage Datasets: View, inspect, and delete your document collections from the "Datasets" tab.

Example Queries

• General: "Summarize the key findings from the Q4 reports."
• Doctor: "What are the patient's pre-existing conditions and current medications?"
• Lawyer: "Find all precedents related to intellectual property theft in the provided case files."
• HR: "What is the company's official policy on remote work and what are the eligibility criteria?"

🆕 New Features

Document Summarization

Navigate to the "Summarize" tab to generate various types of summaries:

• Summary Types: Executive, Key Points, Chapter-wise, Technical, Bullet Points, Abstract
• Length Options: Brief (1-2 paragraphs), Standard (1 page), Detailed (2-3 pages)
• Custom Instructions: Add specific guidance for the summary generation

Interactive Chat

Use the "Chat" tab for conversational document analysis:

• Create chat sessions with one or multiple datasets
• Ask follow-up questions with maintained context
• View citations and sources for each response
• Export conversation history as JSON or Markdown

CSV Analytics

Upload CSV files and analyze them using natural language:

• Example Queries:
  • "What is the average sales by region?"
  • "Show me the trend of revenue over time"
  • "Find correlations between variables"
• Automatic Visualizations: Line charts, bar graphs, heatmaps, scatter plots
• Statistical Analysis: Descriptive statistics, correlations, distributions
• Data Export: Download results as JSON, CSV, or HTML reports

---