docker/openwebui/docker-compose.yml at main · wnstify/docker · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
# Open WebUI — Hardened Docker Compose (single service)
# Self-hosted LLM front-end: https://github.com/open-webui/open-webui
#
# Architecture (1 service):
#   - openwebui  — ghcr.io/open-webui/open-webui (Python/FastAPI app,
#                  bundles SQLite for chat history + ChromaDB for RAG)
#
# Note on dependencies: Open WebUI itself needs nothing external. To
# actually chat, point it at a model provider — either an Ollama
# instance (on the host, in another container, or remote) or an
# external API key (OpenAI, Anthropic, etc.). See README for examples.
#
# Security baseline (verified by test, May 2026):
#   - Runs as ${PUID}:${PGID} (no root). Port 8080 is unprivileged.
#     cap_drop: ALL with ZERO cap_add — verified healthy under test.
#   - security_opt: no-new-privileges; ipc: private
#   - tmpfs for ephemeral writes; deploy.resources limits incl. pids
#   - 127.0.0.1 port binding — reverse proxy in front for TLS
#   - Healthcheck via the built-in /health endpoint
#   - WEBUI_SECRET_KEY pinned via .env so sessions survive restarts
#
# Setup:
#   cp .env.example .env  &&  fill WEBUI_SECRET_KEY (openssl recipe in the file)
#   docker network create openwebui-front
#   mkdir -p data
#   docker compose up -d

name: openwebui

services:

  openwebui:
    image: ghcr.io/open-webui/open-webui:0.9.5
    container_name: openwebui
    restart: unless-stopped
    user: "${PUID:-1000}:${PGID:-1000}"
    ports:
      # localhost-only — front with a reverse proxy for public TLS.
      - "127.0.0.1:${OWUI_PORT:-8071}:8080"
    environment:
      # CRITICAL: signs session cookies. Losing this logs everyone out
      # but doesn't lose data; leaking it lets attackers forge sessions.
      WEBUI_SECRET_KEY: ${WEBUI_SECRET_KEY:?secret key required}
      WEBUI_AUTH: ${WEBUI_AUTH:-true}
      TZ: ${TZ:-Europe/Bratislava}
    # env_file pulls in optional OLLAMA_BASE_URL, OPENAI_API_KEY,
    # OAUTH_* (Authentik SSO), etc. without forcing them to be set.
    env_file:
      - .env
    volumes:
      - ./data:/app/backend/data
    networks:
      - openwebui-front
    cap_drop:
      - ALL
    security_opt:
      - no-new-privileges:true
    ipc: private
    tmpfs:
      - /tmp:size=256m
    deploy:
      resources:
        limits:
          # Open WebUI loads a sentence-transformers embedding model on
          # first boot (~100 MB). Bump if you enable image generation
          # or local RAG with large documents.
          memory: 2g
          cpus: "2.0"
          pids: 300
        reservations:
          memory: 512m
          cpus: "0.25"
    healthcheck:
      test: ["CMD", "curl", "-fsS", "http://127.0.0.1:8080/health"]
      interval: 30s
      timeout: 10s
      retries: 5
      start_period: 90s

# ═══════════════════════════════════════════════════════════════════════════
# NETWORKS
# ═══════════════════════════════════════════════════════════════════════════

networks:
  # External bridge: reverse proxy reaches openwebui via 127.0.0.1 port
  # binding; also gives openwebui internet egress for external model
  # providers (OpenAI, Anthropic) and the embedding-model download.
  openwebui-front:
    external: true