Groundly AI – Multi-Tenant Enterprise RAG Platform

Groundly AI is a production-ready, multi-tenant Retrieval-Augmented Generation (RAG) platform that enables organizations to securely search, retrieve, and interact with internal knowledge using natural language. Built with FastAPI, PostgreSQL (pgvector), Redis, and Google Gemini, the platform allows employees to ask questions across company documents while enforcing tenant, department, and role-based access controls. Groundly combines semantic search, keyword search, reranking, conversational memory, and intelligent agents to deliver accurate, cited, and secure responses.

Features

• Multi-tenant architecture with complete tenant isolation.
• Department and role-based document access.
• Hybrid Retrieval (Vector Search + Full-Text Search).
• Input and output Guardrails
• Cohere reranking for improved retrieval quality.
• AI-powered conversational search using Google Gemini.
• Semantic response caching with RedisVL.
• PDF citations with page references.
• Document ingestion pipeline with asynchronous processing.
• Support for large enterprise knowledge bases.
• Conversation history and persistent chat sessions.
• Background workers powered by ARQ.
• Connectors: connect document from another source
• Observability with Prometheus and Grafana

System Architecture Diagram

System Screenshot

Dashboard

Invite user to department

Department

ChatUI

Connectors

Document Upload

Settings

Tech Stack

• Python
• FastApi
• React
• TenStack Query
• Pgvector
• PostgreSQL
• AsyncPG
• SQLModel
• Redis
• RedisVL
• Guardrail AI
• Google Gemini
• Cohere Reranking
• ARQ

Python Package Manager

• uv

Supported Document

• PDF

Supported Connectors

• DropBox
• Google Drive
• S3 Digital ocean, Aws

Evaluation

Groundly is evaluated using Ragas across a curated dataset. The evaluation report and methodology are available in the evaluation directory.

Use Cases

Groundly is designed for organizations that need secure AI-powered knowledge retrieval, including:

• Insurance
• Banking
• Healthcare
• Legal
• Internal Knowledge Bases
• Enterprise Documentation

infrastructure

• docker
• Docker Compose

Project Structure

Backend/
Frontend/
compose.yml

Project Setup

1. clone Repository

```bash
git clone https://github.com/owolabi-develop/MemRag.git
```

2. Create .env file for both Frontend and backend in root folder

backend .env

• DB_NAME=""
• DB_USER=""
• DB_PASSWORD=""
• DB_HOST=""
• DB_PORT=5432
• REDIS_SERVER="redis"
• REDIS_PORT=6379
• PGADMIN_MAIL=""
• PGADMIN_PW=""
• SECRET_KEY=""
• ALGORITHM="HS256"
• GUARDRAILS=""
• ACCESS_TOKEN_EXPIRE_MINUTES=48
• GOOGLE_CLOUD_LOCATION="us-central1"
• APP_NAME=Groundly
• PASSWORD_RESET_TOKEN_EXPIRE_MINUTES=30
• FRONTEND_HOST="http://localhost" without docker http://localhost:8080
• SMTP_HOST=
• SMTP_PORT=465
• SMTP_TLS=true
• SMTP_SSL=false
• SMTP_USER=
• SMTP_PASSWORD=
• EMAILS_FROM_EMAIL=
• EMAILS_FROM_NAME=Groundly
• INVITE_TOKEN_EXPIRE_MINUTES=4320
• SPACES_SECRET=""
• SPACES_KEY=""
• SPACES_BUCKET_NAME_KB="groundlykb"
• SPACES_ENDPOINT="https://sfo3.digitaloceanspaces.com"
• SPACES_REGION="sfo3"

frontend .env

• VITE_API_URL=http://localhost/api/v1

Running locally

1. Start the backend and frontend services using Docker Compose

docker-compose up --build

Start the backend services

Note: update your database setting on code inside connection.py and db.py

cd backend
uv .venv
uv sync
fastapi dev

start

the frontend services

cd frontend/Mem-rag
npm install
npm run dev

login/signup

• Navigate to http://localhost/login
• Navigate to http://localhost/register

backend APi

• Navigate to http://localhost/api/v1/docs

pgadmin

• Navigate to http://localhost.db

Prometheus

Available on docker

• Navigate to http://promethus.localhost

Grafana

Available on docker

• Navigate to http://grafana.localhost