EPIC.search

EPIC.search is an AI-powered document search and retrieval system that combines vector search capabilities with Large Language Models (LLMs) to provide intelligent document search and question answering capabilities.

Project Overview

The system uses a modern tech stack and architecture:

• React-based web interface
• Python Flask APIs
• Vector search using PostgreSQL with pgvector
• LLM integration using OLLAMA
• Docker-based deployment

Core Components

• search-web: React-based frontend application
• search-api: Flask-based orchestration API
• search-vector-api: Vector search engine API
• search-model: OLLAMA-based LLM service
• embedder: Document processing and embedding tool

Quick Start

Prerequisites

• Docker
• Docker Compose

Running the Stack

1. Clone the repository

git clone <repository-url>
cd EPIC.search

2. Set up environment variables

# Copy sample env files for each component
cp search-api/sample.env search-api/.env
cp search-vector-api/sample.env search-vector-api/.env
cp search-web/sample.env search-web/.env

3. Start the services

docker compose up -d

This will start all the required services:

• Web UI at http://localhost:3000
• Search API at http://localhost:3200
• Vector API at http://localhost:3300
• OLLAMA Model at http://localhost:11434

Component Architecture

sequenceDiagram
    participant Client as Client
    participant WebUI as Web UI
    participant WebAPI as Web API
    participant VectorAPI as Vector API
    participant VectorDB as Vector DB
    participant LLM as LLM Model
    
    Note over WebUI,LLM: All components within Docker network
    
    Client->>WebUI: User query
    WebUI->>WebAPI: Forward query
    WebAPI->>VectorAPI: Process query
    VectorAPI->>VectorDB: Keyword search
    VectorDB-->>VectorAPI: Search results
    VectorAPI->>VectorDB: Semantic search
    VectorDB-->>VectorAPI: Search results
    VectorAPI->>VectorAPI: Rank & combine results
    VectorAPI-->>WebAPI: Return ranked results
    WebAPI->>LLM: Generate response
    LLM-->>WebAPI: RAG response
    WebAPI-->>WebUI: Return response
    WebUI-->>Client: Display to user

Loading

AI Models

Vector Search Models

• Cross Encoder: cross-encoder/ms-marco-MiniLM-L-2-v2
• Embeddings: all-mpnet-base-v2
• Keyword Processing: all-mpnet-base-v2

LLM Configuration

• OLLAMA-based model service
• Default configuration: qwen2.5:0.5b
• Configurable via environment variables:
  • MODEL_NAME: Base model (e.g., qwen2.5, llama2)
  • MODEL_VERSION: Model size/version (e.g., 0.5b, 3b)

Detailed Documentation

Each component has its own detailed documentation:

• Web UI Documentation
• Search API Documentation
• Vector API Documentation
• Document Embedder Documentation

For infrastructure and deployment details:

• Application Architecture
• Infrastructure & Deployment
• System Overview

Development

Running Individual Components

Each component can be run independently for development:

Search API

cd search-api
make setup
make run

Vector API

cd search-vector-api
make setup
make run

Web UI

cd search-web
npm install
npm run dev

Document Embedder

cd tools/embedder
pip install -r requirements.txt
python main.py

Environment Variables

Each component requires specific environment variables. Check the sample.env files in each component directory for the required variables.

Current Status

The project is currently deployed in the BC Gov Landing Zone Test environment. Production deployment is planned for the future.

License

See the LICENSE file for details.