Try Aurora in five minutes (PoC)

Four containers (Postgres, Qdrant, the API and the web app), prebuilt images, nothing to build. Your data stays on your machine.

Install

OS	Command
macOS / Linux	`curl -fsSL https://zyvorai.github.io/aurora/install.sh \	bash`
Windows (PowerShell, Docker Desktop with WSL2)	`irm https://zyvorai.github.io/aurora/install.ps1 \	iex`
Kubernetes	`./scripts/poc/install-k8s.sh` (see below, coming with the Helm PoC chart)
No script, just Docker	download `docker-compose.poc.yml`, then `docker compose -f docker-compose.poc.yml up -d --wait`

You need Docker (Docker Desktop on macOS/Windows, Docker Engine + Compose v2 on Linux) and about 4 GB of free RAM. The installer asks which LLM to use, generates secrets, starts everything and opens http://localhost:3000.

Sign in with marketing@zyvor.dev / Admin@321 (a seeded demo account: change it for anything beyond a throwaway test).

Choose an LLM

Option	Set in `~/.aurora-poc/.env`	Notes
OpenAI-compatible API	`LLM_PROVIDER=openai`, `OPENAI_API_KEY=…` (optional `OPENAI_BASE_URL`, `OPENAI_MODEL`)	Easiest. Embeddings use the provider's embedding model.
Ollama on your computer	`LLM_PROVIDER=ollama`, `OLLAMA_BASE_URL=http://host.docker.internal:11434/v1`	Free and private. Run `ollama pull llama3.1:8b` and `ollama pull nomic-embed-text`; set `OLLAMA_HOST=0.0.0.0` so Docker can reach it. On Apple Silicon Ollama uses the GPU.

Embedding dimensions differ (OpenAI 1536, Ollama 768). Pick one and stick to it; switching later needs aurora-poc reset.

Everyday commands

~/.aurora-poc/aurora-poc <command> (Windows: & "$HOME\.aurora-poc\aurora-poc.ps1" <command>)

Command	What it does
`status`	containers and `/health`
`logs [service]`	follow logs (`api`, `web`, `postgres`, `qdrant`)
`down` / `up`	stop / start (data is kept)
`update`	pull newer images and restart
`llm`	change the LLM settings
`reset`	delete all PoC data and start fresh
`feedback`	open the feedback form pre-filled with your OS
`uninstall`	remove containers, data and the install directory

Data lives in Docker volumes aurora-poc_postgres_data and aurora-poc_qdrant_data. Secrets are in ~/.aurora-poc/.env (mode 600). Change the web port with AURORA_PORT=3100 before installing.

What to try

Create a product from your website or docs URL, then ingest it.
SEO tab: audit a public page, research keywords (estimates are labelled), score a draft.
Content: generate an seo_article, approve it, publish (WordPress/Dev.to/… need credentials).

Troubleshooting

Port 3000 busy: AURORA_PORT=3100 aurora-poc install.
"Docker is not running": start Docker Desktop and retry.
Agents say the LLM is not ready: run aurora-poc llm, and check aurora-poc status shows llm_ready: true.
Windows: the script is written and linted but has not been run on a real Windows machine yet; please report anything odd.
Apple Silicon / ARM Linux: images are multi-arch (amd64 + arm64).

Tell us how it went

aurora-poc feedback opens a short form (ease of install, what worked, what hurt). Or open an issue.

Under the hood

docker-compose.poc.yml runs DEPLOYMENT_PROFILE=minimal (no Neo4j, Redis workers or MinIO; artifacts live in Postgres and workflows run in the API process). The web container proxies /api/v1 to the API at runtime (API_INTERNAL_URL), so the same prebuilt image works behind any host name. For production topologies see docs/dev-guide.md and k8s/README.md.

Edit this page on GitHub