f5c756707f · krz/brand-bench

Dockerfile added +18

		@@ -0,0 +1,18 @@
	1	# ── Stage 1: build ───────────────────────────────────────────────────────────
	2	FROM node:22-alpine AS build
	3
	4	WORKDIR /app
	5
	6	COPY package*.json ./
	7	RUN npm ci
	8
	9	COPY . .
	10	RUN npm run build
	11
	12	# ── Stage 2: serve ───────────────────────────────────────────────────────────
	13	FROM nginx:alpine
	14
	15	COPY --from=build /app/dist /usr/share/nginx/html
	16	COPY docker/nginx.conf /etc/nginx/conf.d/default.conf
	17
	18	EXPOSE 80

README.md +26

		@@ -74,6 +74,32 @@ Larger models produce better-structured output. If generation fails with a JSON
74	74	error, try a bigger model. Generation typically takes 15–60 seconds depending on
75	75	hardware.
76	76
	77	## Self-hosting with Docker
	78
	79	A `Dockerfile` and `compose.yml` are included for deploying the app alongside
	80	Ollama on a server.
	81
	82	```bash
	83	docker compose up -d --build
	84	```
	85
	86	The app is served on port 8000. Ollama's API is proxied through nginx at `/api/`
	87	so the browser never makes a cross-origin request — no CORS configuration
	88	needed.
	89
	90	After first boot, pull at least one model:
	91
	92	```bash
	93	docker compose exec ollama ollama pull llama3.2
	94	```
	95
	96	Then open the app, go to Settings (⚙), enable AI, and set the Ollama base URL
	97	to `http://your-server:8000` (no path suffix).
	98
	99	GPU support: If your server has an NVIDIA GPU, uncomment the `deploy` block
	100	in `compose.yml` (requires the
	101	[NVIDIA Container Toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html)).
	102
77	103	## Project structure
78	104
79	105	```

compose.yml added +35

		@@ -0,0 +1,35 @@
	1	services:
	2
	3	app:
	4	build: .
	5	ports:
	6	- "8000:80"
	7	depends_on:
	8	- ollama
	9	restart: unless-stopped
	10
	11	ollama:
	12	image: ollama/ollama
	13	volumes:
	14	- ollama_data:/root/.ollama
	15	restart: unless-stopped
	16	# Expose port 11434 if you want to reach Ollama directly from your host
	17	# (e.g. to run `ollama pull` without exec-ing into the container).
	18	ports:
	19	- "11434:11434"
	20
	21	# ── GPU support ──────────────────────────────────────────────────────────
	22	# Uncomment the block below if your server has an NVIDIA GPU.
	23	# Requires the NVIDIA Container Toolkit:
	24	# https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html
	25	#
	26	# deploy:
	27	# resources:
	28	# reservations:
	29	# devices:
	30	# - driver: nvidia
	31	# count: all
	32	# capabilities: [gpu]
	33
	34	volumes:
	35	ollama_data:

docker/nginx.conf added +27

		@@ -0,0 +1,27 @@
	1	server {
	2	listen 80;
	3	root /usr/share/nginx/html;
	4	index index.html;
	5
	6	# SPA fallback — let React Router handle unknown paths
	7	location / {
	8	try_files $uri $uri/ /index.html;
	9	}
	10
	11	# Proxy Ollama's API through the same origin so the browser never has to
	12	# make a cross-origin request and CORS is a non-issue.
	13	# In the app's Settings panel, set the Ollama base URL to just:
	14	# http://<your-server> (no port, no path suffix)
	15	location /api/ {
	16	proxy_pass http://ollama:11434/api/;
	17	proxy_http_version 1.1;
	18	proxy_set_header Host $host;
	19	proxy_set_header X-Real-IP $remote_addr;
	20	proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
	21
	22	# Ollama responses (especially streamed ones) can be large and slow;
	23	# disable buffering so cancellation works promptly.
	24	proxy_buffering off;
	25	proxy_read_timeout 120s;
	26	}
	27	}

Commit f5c756707f