Skip to main content

AI Chat

The AI Chat is a conversational interface that sits next to the planning view and lets users ask questions about the schedule, run quick actions, and let the assistant call Dime.Scheduler tools on their behalf. Under the hood, the chat connects an LLM provider to the same toolset the MCP server exposes - which means it can search tasks, look up resources, find available slots, create or reschedule appointments, send messages, and run route optimization.

Enabling the chat​

The chat is off by default. To enable it, open Application Setup, switch to the Preview tab and configure the AI Chat section:

  1. Toggle Show AI chat on.
  2. Pick a Provider: OpenAI, Anthropic, or Azure OpenAI.
  3. Enter a Model identifier (e.g. gpt-4o-mini, claude-haiku-4-5-20251001). Leave blank to use the provider default.
  4. Enter the API key for the selected provider.
  5. For Azure OpenAI, also fill in the Endpoint (e.g. https://<resource>.openai.azure.com) and the Deployment name.
  6. Save.

Until an API key is configured, the chat panel shows a configuration prompt with a link back to the setup page rather than the conversation UI.

caution

The API key is stored at tenant level and is used for every user's chat session. Make sure to use a key from a dedicated project or budget so usage can be monitored.

Conversation surface​

Once configured, the chat shows three regions:

  • Sidebar with all saved conversations. Clicking a conversation reopens it; the New conversation button starts a fresh thread.
  • Main window with the active conversation, message input, and a header with the title and description.
  • Side actions: Quick Actions, Examples, and Download.

Quick Actions​

Quick Actions are pre-built workflows that bypass the free-form chat. They show forms with the parameters the underlying tool expects, validate the input, and call the tool with the result. Use them when you know exactly which tool you want to run and want to skip the conversational round-trip.

Examples​

The Examples panel lists capability cards with sample questions for each tool category. Clicking an example pre-fills the message input so you can tweak it before sending.

Download​

The download button exports the current conversation as a text file. Useful for sharing context with colleagues or attaching to a support ticket.

What the chat can do​

The chat uses the in-product MCP toolset, so anything documented under Model Context Protocol is available conversationally. Common patterns:

  • "What's open for this week?" - calls the open-tasks resource.
  • "Find a 2-hour slot for John Smith next Monday." - uses find_available_slots.
  • "Move tomorrow's 14:00 appointment for Jane to 15:00." - uses reschedule_appointment.
  • "Optimize routes for John, Jane and Lara from Monday to Friday." - uses optimize_field_service and explains the trade-offs.
  • "Notify John his 14:00 moved to 15:00." - uses send_message.

When a tool call could change data (creating, rescheduling, or deleting appointments, or sending notifications), the chat asks for confirmation before it executes.

Read more​

Last updated on by Hendrik Bulens