LibreChat Setup Without Docker Compose

LibreChat Setup Without Docker Compose | Generated by AI

Home 2025.10

Yes, you can set up LibreChat without Docker Compose by using individual docker run commands or even a non-Docker method (manual installation with Node.js, Yarn, MongoDB, and MeiliSearch). This avoids downloading the docker-compose binary itself (which is small, ~20-30MB, but I assume you’re referring to the overall image pulls totaling ~70MB+ for LibreChat’s services). However, the core Docker images (e.g., for the app, MongoDB, MeiliSearch) are still around 500MB-1GB total when pulled—Docker Compose is just an orchestration tool on top of Docker, not the main bandwidth user. Pulling images over slow 4G/5G will still be the bottleneck, but you can mitigate it.

I’ll outline options below, prioritizing bandwidth-saving tips first. If mobile data is very limited, consider tethering to a WiFi network temporarily or using a pre-downloaded setup on another machine (e.g., export/import images via docker save/docker load).

Bandwidth-Saving Tips for Any Docker-Based Setup

Pre-pull images on a faster connection: On another device with better internet, run docker pull node:20-alpine (for the app), docker pull mongo:7 (database), and docker pull getmeili/meilisearch:v1.10 (search). Then save them to files:
```
docker save -o librechat-app.tar node:20-alpine
docker save -o mongo.tar mongo:7
docker save -o meili.tar getmeili/meilisearch:v1.10
```
Transfer the .tar files via USB/drive (total ~500-800MB compressed), then on your Ubuntu machine: docker load -i librechat-app.tar etc. No online pulling needed.
Use lighter alternatives: For testing, skip MeiliSearch initially (it’s optional for search; disable in config). MongoDB image is ~400MB—use a local MongoDB install instead (see non-Docker section below) to save ~400MB.
Monitor usage: Use docker pull --quiet or tools like watch docker images to track.
Proxy or cache: If you have a Docker Hub mirror or proxy, configure it in /etc/docker/daemon.json to speed up pulls.

Option 1: Pure Docker (No Compose) – Equivalent to Compose Setup

You can replicate the docker-compose.yml behavior with docker run and docker network. This downloads the same images but lets you control each step. Total download still ~700MB+ (app build + DBs).

Create a Docker network (isolates services):
```
docker network create librechat-network
```

Run MongoDB (replace your_mongo_key with a strong password):

docker run -d --name mongodb --network librechat-network \
  -e MONGO_INITDB_ROOT_USERNAME=librechat \
  -e MONGO_INITDB_ROOT_PASSWORD=your_mongo_key \
  -v $(pwd)/data/mongodb:/data/db \
  mongo:7

Creates ./data/mongodb for persistence.

Run MeiliSearch (replace your_meili_key):

docker run -d --name meilisearch --network librechat-network \
  -e MEILI_MASTER_KEY=your_meili_key \
  -p 7700:7700 \
  -v $(pwd)/data/meili:/meili_data \
  getmeili/meilisearch:v1.10

Skip if bandwidth is tight; disable in app config later.

Clone and Build/Run the LibreChat App:
- Clone repo if not done: git clone https://github.com/danny-avila/LibreChat.git && cd LibreChat (~50MB download for repo).
- Build the image (this pulls Node.js base ~200MB and builds app layers):
```
docker build -t librechat-app .
```
- Run it (links to DB, uses env vars—create a .env file as in my previous response):
```
docker run -d --name librechat --network librechat-network -p 3080:3080 \
  --env-file .env \
  -v $(pwd):/app \
  librechat-app
```
  - In .env, set MONGODB_URI=mongodb://librechat:your_mongo_key@mongodb:27017/LibreChat and MEILI_HOST=http://meilisearch:7700 etc.
Access: http://localhost:3080. Logs: docker logs -f librechat.

Stop/Cleanup: docker stop mongodb meilisearch librechat && docker rm them.
Pros/Cons: Same as Compose, but more manual. Still heavy on data for image pulls/build.

Option 2: Non-Docker Installation (Manual, No Image Pulls) – Recommended for Low Bandwidth

Install dependencies natively on Ubuntu. This avoids all Docker overhead (~0MB for containers; just package downloads via apt/yarn, totaling ~200-300MB). Uses your system’s Python/Node setups indirectly.

Prerequisites (Install Once)

sudo apt update
sudo apt install nodejs npm mongodb-org redis meilisearch git curl build-essential python3-pip -y  # MongoDB official pkg; MeiliSearch ~50MB binary
sudo systemctl start mongod redis meilisearch
sudo systemctl enable mongod redis meilisearch

Node.js: If not v20+, install via nvm: curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.7/install.sh | bash, then nvm install 20.
Yarn: npm install -g yarn.
MongoDB config: Edit /etc/mongod.conf to bind to localhost, restart.

Installation Steps

Clone Repo:

cd ~/projects
git clone https://github.com/danny-avila/LibreChat.git
cd LibreChat

Install Dependencies:

yarn install  # ~100-200MB downloads for packages

Configure .env (copy from .env.example):
- cp .env.example .env && nano .env
- Key changes:
  - Mongo: MONGODB_URI=mongodb://localhost:27017/LibreChat (create DB user if needed via mongo shell).
  - Meili: MEILI_HOST=http://localhost:7700 and MEILI_MASTER_KEY=your_key.
  - Disable search if skipping Meili: SEARCH=false.
  - Add AI keys as needed.
Build and Run:
- In one terminal: yarn run backend (starts API on port 3090).
- In another: yarn run frontend (starts UI on port 3080).
- Or use PM2 for production: yarn global add pm2 && pm2 start yarn --name backend -- run backend etc.
Access: http://localhost:3080. Create account and configure.

Data: MongoDB stores in /var/lib/mongodb, Meili in /var/lib/meilisearch.
Troubleshooting:
- Mongo auth: Run mongo and create user: use LibreChat; db.createUser({user: 'librechat', pwd: 'yourpass', roles: ['readWrite']}).
- Ports/Firewall: sudo ufw allow 3080,3090,7700.
- Updates: git pull && yarn install && restart services.
Pros: Lighter on bandwidth (no containers), uses system resources directly. Cons: More setup fiddling, potential dependency conflicts (e.g., with your Python 3.13, but Node is separate).

If mobile data is extremely limited, the non-Docker method is best—do it in stages, monitoring with du -sh on directories. For errors, share logs (e.g., from yarn run backend). Official non-Docker guide: https://docs.librechat.ai/install/installation/manual_install.html. If you need Ollama integration for local AI (saves API costs), install it separately via their script (~100MB). Let me know for tweaks!

Back

x-ai/grok-4-fast:free

Donate