Ovrion Get the image

Slack · Teams · Outlook

Workplace AI that never leaves your infrastructure.

Ovrion answers Slack messages, Teams chats, and Outlook drafts using a model you run — on your own hardware, behind your own firewall. Nothing is proxied through us. There is no us in the request path at all.

docker pull & run See the integrations
0 bytes
of prompt content leave your network
Draft-only
Outlook path — no send-mail method exists in the code
Auditable
egress audit endpoint, not a marketing claim

How it works

One image. Your model. Your network.

Ovrion is a container you run yourself, in front of any OpenAI-compatible backend you already have — LM Studio, Ollama, vLLM, llama.cpp server. It coordinates workers, exposes a gateway, and answers workplace webhooks locally.

  1. Pull the imageOne published artifact: ghcr.io/finsavvyai/ovrion, scanned and tagged per release.
  2. Point it at your modelAny OpenAI-compatible endpoint on your network — the proxy worker forwards /v1/chat/completions to it.
  3. Wire up a webhookRegister the Slack/Teams request URL, or call the Outlook endpoints directly — all answered by your gateway, on your box.
  4. Check the audit trail/sovriq/audit/egress reports whether prompt content ever left the perimeter. It's a live endpoint, not a slide.

Integrations

Meet your team where it already works.

Every integration is answered by your own gateway. No integration in this list has a code path that sends a message on your behalf without you calling it explicitly.

💬

Slack

Mentions, DMs, slash commands, and interactive buttons — all HMAC-verified against your signing secret. Bot messages are ignored automatically so it never talks to itself.

🟦

Microsoft Teams

Plain replies or Adaptive Cards, bot-added greetings, and automatic @mention stripping — authenticated with a shared webhook secret you control.

✉️

Outlook

Draft a reply, summarize a thread, or leave a tentative RSVP note via Microsoft Graph — every path is read-or-draft only. Sending mail on your behalf is not implemented, anywhere.

Pricing

Free on your laptop. Paid when you want more.

The cluster itself is free and open-source forever. A subscription only unlocks the Slack/Teams/Outlook integrations layer — and if your own hardware isn't enough, GPU credit gets you real cloud power billed by the second, not the token.

Free

$0

The core LunaCluster cluster, forever. Bring your own model, your own hardware. No integrations layer.

Pro

$29/mo

Everything in Free, plus Slack, Teams, and Outlook — one gateway instance.

Business

$99/mo

Everything in Pro across unlimited gateway instances, plus priority support.

GPU credit (pay-as-you-go)

Want more power than your laptop can give you? Pick any Hugging Face model — Ovrion sizes it and picks the right GPU automatically — and pay our RunPod cost plus a fixed fee, debited by the second while it's warm, out of credit you buy in advance. No tokens, no weekly cap, stop anytime. Every debit is logged to an audit ledger you can pull at any time. See docs/guides/GPU_CREDIT_WALLET.md.

Deploy

Pull it. Run it. It's yours.

No account, no signup, no telemetry callback. The full setup — including wiring up Slack/Teams/Outlook — is documented in the repo.

# pull the released image
docker pull ghcr.io/finsavvyai/ovrion:latest

# run master, worker, and gateway (see docs/guides/DOCKER_IMAGE.md)
docker run -d --name lunacluster-gateway \
  -p 8080:8080 \
  ghcr.io/finsavvyai/ovrion:latest \
  python3 -m src.api.gateway --host 0.0.0.0 --port 8080

# verify it's actually running locally
curl http://localhost:8080/health