Today we're releasing WeeBie Monitor 1.0.0 — our first production version of the self-hosted AI firewall. It's the governed gateway that sits between your people, your applications, and every AI model they use: a single place to monitor, control, and protect every AI interaction in your organization. And it runs entirely on your own infrastructure — cloud, on-premises, or fully air-gapped.
Enterprises stopped experimenting with AI a while ago. They're running copilots, support bots, document pipelines, and autonomous agents that touch sensitive data every day. Each of those interactions leaves the corporate perimeter to reach a model endpoint. WeeBie Monitor brings that interaction back inside your governance envelope — without slowing your teams down.
One line to integrate
Any OpenAI- or Anthropic-compatible app or SDK points its base_url at WeeBie and keeps working. From that moment, every request is identified, policy-checked, PII-scanned, cost-metered, and written to a tamper-evident audit trail — with a single dashboard for observability, cost, alerts, approvals, and prompt versioning. No rewrites, no per-app agents.
What 1.0.0 delivers
1.0.0 is a complete AI governance control plane, not a proof of concept. Out of the box:
- Guardrails & DLP — detect and redact or block PII, secrets, source code, and disallowed content on the way in and the way out.
- AI threat protection — prompt-injection and jailbreak detection, secret-leakage and data-exfiltration checks, and content moderation, on prompts and responses.
- Cost control & FinOps — real-time token metering, per-team budgets and rate limits, chargeback by cost center, anomaly alerts, and hard quotas.
- Tamper-evident audit & compliance — a hash-chained record of every AI interaction, plus one-click evidence reports mapped to SOC 2, GDPR, and the EU AI Act.
- Enterprise IAM — SCIM provisioning, SAML sign-on, fine-grained custom RBAC, and access-key lifecycle with rotation.
- Resilience & reach — automatic provider failover with retries and a circuit breaker, and native routing to OpenAI, Anthropic, Azure OpenAI, AWS Bedrock, Google Vertex, and local models (Ollama / vLLM).
- SIEM export & data governance — stream events to Splunk, Datadog, or Elastic; configurable retention, GDPR erasure, and data-residency routing.
- Developer ecosystem — thin SDKs for Python, JavaScript, Go, Java, and .NET, a Terraform provider, and LangChain / LlamaIndex integrations.
Two editions — including a true air-gap
WeeBie Monitor 1.0.0 ships in two editions. The standard edition installs on any internet-connected Linux server and pulls its version-pinned components at install time. The air-gapped edition is fully self-contained — every container image and a local language model are included, so it installs and runs on a machine that has never touched the internet.
That air-gap path isn't a claim on a slide. We drilled it end to end on a machine with its network physically disconnected: load the images, restore the model, start the stack, and complete a real AI chat through the local model — with the egress lock actively blocking any outbound call. For regulated, classified, and isolated environments, this is the difference between "AI is off-limits" and "AI, safely, on our terms."
One command to install
Deployment is a single command. The installer generates every secret, writes the configuration, starts the stack, installs nightly backups, and runs an acceptance self-test — then prints your dashboard URL and admin login. On a fresh server it's up in minutes, and the database it creates is empty: your data, and only your data, from the first request.
Self-hosted, permissively licensed, yours
WeeBie Monitor runs on your hardware. Prompts, responses, costs, and audit records never leave your network unless you configure an outbound provider — and in air-gap mode they can't leave at all. Every bundled open-source component is permissively licensed for clean commercial use, and the dashboard is white-labelable to your brand. There's no per-seat telemetry phoning home and no lock-in to a single model provider.
Try it today
The fastest way to understand WeeBie is to watch a request flow through it. Our live demo is open — no signup — so you can inspect the policy engine and see guardrails, DLP, cost metering, and the audit trail in real time. When you're ready to run it yourself, the documentation walks through both the standard and air-gapped installs step by step.
This is 1.0.0 — the foundation. It's the version we were comfortable putting a number on, and it's just the start of what's ahead. We'd love your feedback.