Is web search enabled just by installing Open WebUI?

Do not assume so. Configure a supported search provider and test retrieval separately from chat.

Can I use it with a remote model instead?

Yes. Configure a supported hosted or OpenAI-compatible endpoint. Token costs and data handling then depend on that provider.

Why keep WEBUI_SECRET_KEY stable?

It signs sessions. Changing it on container recreation can invalidate sessions; keep it private and persistent.

Open WebUI: a workspace for local models and teams

Core features

Open WebUI is a self-hosted model workspace with Ollama and OpenAI-compatible connections, document workflows and user management. Web search is an integration you configure, rather than a reason to assume every installation behaves like Perplexity.

Keep several models accessible from a shared interface.
Use local document retrieval with suitable embedding configuration.
Set up accounts and review permissions before sharing access.
Add a search provider when live web information is necessary.

Hardware and model requirements

Our 2 GB app allowance is an estimate for a small setup, not an official minimum. It excludes chat-model weights, operating-system memory and workload spikes. The standard image includes local ML components; document indexing can need more memory than idle chat. CPU inference is possible but can be slow. With Ollama, download a chat model and configure embedding separately.

A five-minute Docker path

Prerequisites: Docker Compose, a running Ollama server, and a downloaded model. Time for downloads is additional. Create a persistent signing secret in the deployment directory:

umask 077
printf 'WEBUI_SECRET_KEY=%s\n' "$(openssl rand -hex 32)" > .env

Do not rerun this command over an existing .env without preserving its settings. Save as compose.yml:

services:
  open-webui:
    image: ghcr.io/open-webui/open-webui:main
    ports:
      - "127.0.0.1:3000:8080"
    extra_hosts:
      - "host.docker.internal:host-gateway"
    environment:
      WEBUI_SECRET_KEY: ${WEBUI_SECRET_KEY:?Set WEBUI_SECRET_KEY in .env}
      OLLAMA_BASE_URL: http://host.docker.internal:11434
    volumes:
      - open-webui-data:/app/backend/data
    restart: unless-stopped
volumes:
  open-webui-data:

docker compose config --quiet
docker compose up -d

Open http://localhost:3000 and complete initial account setup before making the service reachable by others. Confirm your Ollama models appear and test a prompt. On Linux, a host Ollama service may need to listen on an interface reachable from Docker; allow only the needed local/container traffic. Download Compose.

The main image tag is mutable. Reference: official quick start checked 2026-10-22 and the pinned repository source below. Record the image digest for rollback.

Strengths and trade-offs

Good fit: local model experimentation, documents and a shared workspace. Trade-offs: web search needs configuration, permissions need review, and model quality remains your responsibility. Back up the persistent data volume and keep the signing secret stable across recreation.

License matters

Open WebUI uses its own license with branding conditions. Do not label the current release as plain MIT or assume all rebranding is permitted. Read the current license for your intended deployment.

How it differs from Perplexity

Open WebUI starts as a general model workspace. Perplexity starts as a hosted answer service. If cited web research is the priority, compare your configured search pipeline against Vane; if model and user management come first, Open WebUI is the more natural starting point.