Self-host

Oracus.ai is self-hosted: it runs entirely inside your network, and your tickets, specs, embeddings, and chat history never leave it. You bring the cloud, the model keys, and the data.

Oracus.ai is a proprietary product, not an open-source project — you don’t get the source. We distribute a pre-built container image, and access is currently granted by request while we onboard early customers.

What you’ll run

  • One Oracus.ai container image (it runs both the API server and the embedding worker — they differ only by entrypoint).
  • A Postgres 16 instance with the pgvector extension.
  • Docker + Docker Compose (Compose v2). Everything ships as a single Compose file.

What you’ll need

  • A host with Docker (Linux, or Docker Desktop on Mac/Windows).
  • An OpenAI API key — used for the 1024-dim embeddings (text-embedding-3-large with Matryoshka truncation).
  • Optionally an Anthropic API key for the generation tasks (bug_coverage, qa_chat, semantic_link).

No data and no API keys ever reach us — you configure providers in your own instance, and your keys are encrypted at rest.

Request access

Self-host access is invitation-based right now. Tell us a bit about your team and we’ll send you:

  1. A read-only token to pull the Oracus.ai image.
  2. A ready-to-run docker-compose.prod.yml.
  3. A .env template and a 5-minute first-run walkthrough.

Request self-host access

Once you’re approved, bringing Oracus.ai up is a single docker compose up -d — Postgres comes up first, then the app runs its migrations and opens on port 3000.