Chímaira /kaɪˈmɪərə/

Agentic RAG (Retrieval-Augmented Generation) Web Application

Overview

Chímaira is a RAG (Retrieval-Augmented Generation) architecture based web application designed to provide an autheticated end user with an AI augmented productivity tool for managing reusable contexts and generative AI responses (RAG Artifacts) in tandem with accomplishing specialized tasks via AI Agents; cleverly utilizing Langchain framework and Huggingface: Transformer & Pipelines to facilitate a user-friendly integration of local, edge and cloud hosted large language models (LLMs) with the extensible N8N workflow automation platform for agentic response actions.

Related Links

Portfolio URL >_ https://www.mekaegwim.ca/portfolio/ Demo URL >_ TBD

Prerequisites/System Requirements

> NodeJS 20+
> Docker Desktop (for local container deployment)
> PostgreSQL Database server (for non SQLite data persistence)
> OpenAI, Anthropic, xAI API service account/key

Option 1: Docker Container

docker compose up conjureschimera

Option 2: Local

npm install && npm install -g nodemon

# Development build
npm run dev

# or Production build
npm install pm2
npm run start

System Design & Architecture

Synopsis

• Neumorphism Design UI Components
• Passport.js JWT User Authentication & Role Based Access Control
• Nodemailer SMTP Transport - Email Notification(s)
• Agentic RAG Request roles - Explorer, Translator, Historian, Engineer, Artificer, Treasurer
• Reusable Context Profiles
• Dynamic Context Query Models (Query Models)
• Langchain Embedding Models & Vector Database Integration
• Open AI, Anthropic, xAI, Github & N8N Workflow Automation API Integration

Architecture

Toolstack

JavaScript, Typescript, Material UI, ReactJS, Axios, Redux, Bcrypt, JWT, Passport.js, Access Control, Formik, Yup, Luxon, Express, Nodemailer, Mailgen, Sequelize, SQLite, PostgreSQL, Pinecone, ChromaDB, Langchain, Huggingface, N8N, OpenAI, xAI and Anthropic LLM APIs.