On this page
Key takeaways
- Run DashboardGPT on the AI model your organisation chooses, including one on your own servers, so questions about your data never leave your control.
- Ollama, vLLM, LM Studio and company AI gateways all connect through one simple "OpenAI compatible" option in Settings.
- A built-in connection test confirms everything works before you save, so your team gets a smooth first experience.
- Pair a private model with an on-premise install and the whole analytics experience stays inside your walls.
Why teams run their own model
DashboardGPT works on a bring-your-own-key basis: you decide which AI provider answers your team's questions. Many teams pick OpenAI, Anthropic, Google Gemini or Azure OpenAI. Others want the model on their own hardware, and DashboardGPT supports that just as easily.
- Your data stays with you. Questions and the context behind them go to a model you run, not to an outside service.
- Meet policy and contract requirements. If your organisation cannot send data to an external AI provider, you can still give every team plain-English analytics.
- Predictable costs. A model on your own hardware does not charge per question, so heavy use does not mean a surprise bill.
- Freedom of choice. Use an open model you trust, or route through a company AI gateway your IT team already manages.
Works with the tools you already use
If your team already runs models with Ollama, vLLM or LM Studio, or offers AI through an internal gateway, DashboardGPT can use it. All of these connect through a single option called OpenAI compatible. There is nothing to install on the DashboardGPT side and no custom integration work.
Everything else stays the same for your users: they ask questions, get charts and answers they can check, and build dashboards on a shared canvas. Only the engine behind the answers changes.
Connect your model in a few clicks
An organisation owner or administrator sets this up once for the whole team:
- Open Settings, go to the AI tab and choose Add provider.
- Pick OpenAI compatible and give it a friendly name, such as "Our in-house model".
- Enter the address your IT team gives you for the model, and the model name.
- Add an access key only if your model asks for one.
- Choose Test connection, then save.
Any key you add is stored securely and never shown in full again. Your team can start asking questions straight away.
Confidence before you roll out
The connection test checks that your model is reachable and answering before you save, so problems are caught by the administrator rather than discovered by the team. If anything needs attention, you see a clear message explaining what to fix.
Your choice is respected. When you connect your own model, DashboardGPT uses that model. Your questions are never quietly sent somewhere else.
A good rollout tip: try a handful of your team's real questions with your chosen model first. Larger models tend to handle complex analysis best, so pick one that fits the questions your team asks most.
Want DashboardGPT itself on your own infrastructure too? Read the on-premise overview and the documented system requirements, compare plans and pricing, or talk to us about your setup. Ready to start now? Create your free account.