Concious

Concious is a full-stack second-brain app for collecting useful links, organizing them in a personal dashboard, sharing a public version of your brain, and querying saved content through hybrid search and a RAG-powered chatbot.

Highlights

• Password-based signup and signin
• JWT-protected dashboard
• Create, edit, delete, and list saved content
• Shareable public brain link
• Responsive landing page and dashboard experience
• Hybrid search:
  • lexical keyword matching for direct hits
  • semantic retrieval for meaning-based discovery
• RAG chatbot (Ashqnor) grounded on the user's saved content
• MongoDB Atlas Vector Search over stored content embeddings

Tech Stack

• Frontend: React 19, TypeScript, Vite, Tailwind CSS, React Router, TanStack Query, Axios, GSAP
• Backend: Node.js, Express 5, TypeScript, MongoDB Atlas, Mongoose, JWT, bcryptjs, Zod
• AI / Retrieval:
  • Hugging Face Inference API for embeddings
  • MongoDB Atlas Vector Search for semantic retrieval
  • OpenRouter for response generation
• Architecture: hybrid search + Retrieval-Augmented Generation (RAG)
• Tooling: npm workspaces, ESLint, TypeScript project builds

AI Search And Chat

Concious now includes a retrieval layer on top of saved content:

1. A saved content item is embedded and stored with an embedding vector.
2. MongoDB Atlas Vector Search indexes the embedding field.
3. Search combines:
   • lexical matching for obvious keyword hits
   • semantic similarity for related meaning
4. Chat uses RAG:
   • embed the user question
   • retrieve relevant saved items
   • pass grounded context to the model
   • return a response with source references

This keeps normal word search useful while improving recall with semantic retrieval.

Project Structure

.
|-- concious_backend/      # Express + TypeScript API
|   |-- src/
|   |-- .env.example
|   `-- package.json
|-- concious_frontend/     # React + Vite client
|   |-- public/
|   |-- src/
|   |-- .env.example
|   `-- package.json
|-- .gitignore
|-- package.json
`-- README.md

Getting Started

Prerequisites

• Node.js 20 or newer
• npm 10 or newer
• A MongoDB Atlas connection string
• A Hugging Face API token with inference provider access
• An OpenRouter API key

1. Install dependencies

npm install

2. Configure the backend

cd concious_backend
cp .env.example .env

Update concious_backend/.env:

PORT=3000
MONGO_URI=mongodb+srv://<user>:<password>@<cluster>/<db>
JWT_PASSWORD=replace-with-a-long-random-secret
HF_API_KEY=hf_xxxxxxxxxxxxxxxxx
HF_EMBEDDING_MODEL=intfloat/e5-small-v2
OPENROUTER_API_KEY=sk-or-v1-xxxxxxxxxxxxxxxxx
OPENROUTER_MODEL=mistralai/mistral-small-3.1-24b-instruct

3. Create the MongoDB Atlas Vector Search index

Create a vector index on the contents collection with:

• Index name: vector_idx
• Field: embedding
• Dimensions: 384
• Similarity: cosine

Atlas index definition:

{
  "fields": [
    {
      "type": "vector",
      "path": "embedding",
      "numDimensions": 384,
      "similarity": "cosine"
    }
  ]
}

4. Configure the frontend

cd ../concious_frontend
cp .env.example .env

Update concious_frontend/.env if your backend runs somewhere else:

VITE_BACKEND_URL=http://localhost:3000

5. Run the app

Open two terminals:

npm run dev:backend

npm run dev:frontend

The frontend will run through Vite. The backend listens on the PORT from concious_backend/.env.

Available Scripts

From the repository root:

npm run dev:backend
npm run dev:frontend
npm run build
npm run build:backend
npm run build:frontend
npm run lint:frontend

API Overview

Method	Route	Purpose
POST	/api/v1/signup	Create an account
POST	/api/v1/signin	Sign in and receive a JWT
GET	/api/v1/content	List authenticated user's content
POST	/api/v1/content	Add content
PATCH	/api/v1/content/:id	Rename content
DELETE	/api/v1/content/:id	Delete content
POST	/api/v1/search	Hybrid lexical + semantic search
POST	/api/v1/chat	RAG chatbot over saved content
POST	/api/v1/reindex-embeddings	Backfill embeddings for existing content
POST	/api/v1/brain/share	Create or remove a share link
GET	/api/v1/brain/:shareLink	Read a shared public brain

How Search Works

Concious does not replace normal search with semantic search. It enhances it.

• Keyword search still returns strong direct matches such as exact titles or obvious terms.
• Semantic search improves retrieval when the wording differs but the meaning is related.
• The backend merges both result sets so users keep predictable search behavior while benefiting from embeddings.

Security Notes

• Keep real .env files out of Git.
• Set a strong JWT_PASSWORD in every deployed environment.
• Do not commit MongoDB usernames, passwords, or connection strings.
• Rotate Hugging Face and OpenRouter keys if they are ever exposed.