What the assistant does
The assistant is the chat in the app (web app: AI Assistant in the sidebar; mobile app: the Assistant tab). It can set up everything a member can set up in the app: monitors and flows, scheduled tasks, notification channels, status pages, incidents, log alerts and variables. It asks before it deletes or publishes anything.
It sees configuration and check results, not your application's content: of a check it gets the status, status code, error message, timings and the failed assertion, never the response body or headers, and never the value of a secure variable.
In the mobile app you can ask it from a failing monitor, from the log search and from an outage notification (Why?); the conversation then starts with that monitor, search or outage.
Log access
With Log access on, the assistant can search your logs (UptimeEye Logs) to answer questions like "why was the shop down at 3 am?": it counts lines over time, looks at the most frequent values of a field, reads matching lines and the context around one entry, and previews log alerts.
- Open the assistant and the settings (the gear in the chat header; in the mobile app the settings icon)
- Switch on Log access (organization admins only)
Own model provider
An admin can run the assistant on a provider of the organization's choice, for example one you already have a data processing agreement with. Every endpoint that speaks the OpenAI Chat Completions API with tool calling works: OpenAI, Azure OpenAI, Mistral, Anthropic's OpenAI-compatible API, or a self-hosted model behind vLLM or Ollama.
| Provider | Base URL | Example model |
|---|---|---|
| OpenAI | https://api.openai.com/v1 | gpt-4.1 |
| Azure OpenAI | https://<resource>.openai.azure.com/openai/v1 | your deployment name |
| Mistral | https://api.mistral.ai/v1 | mistral-large-latest |
| Anthropic | https://api.anthropic.com/v1 | claude-sonnet-5-5 |
- Open the assistant settings in the web app
- Pick a provider, check the base URL, enter the model and the API key
- Click Save and test: UptimeEye makes one test call that must call a tool and shows the result
- The API key is stored encrypted and never shown again, only its last four characters.
- Every conversation of the organization then runs on your provider. If it fails, the assistant shows the provider's error; it never falls back to UptimeEye's provider.
- With your own provider, Log access is switched on (you can switch it off); removing the provider switches it off again.
- The assistant chooses from about 70 tools. Small models often fail at that; the test call tells you early.