The platform in practice

Slimaco assistant

The main interface for working with Slimaco — questions, answers with citations and controlled access to tools.

How an answer is built

1. Receiving the question — text, attachments, an assistant reference and a source scope.
2. Access control — permissions, plan, sources and tools available to the user.
3. Choosing the execution mode — standard mode, an assistant, a multi-assistant plan, a skill, or a tool.
4. Retrieving knowledge — searching for evidence within the approved and permitted scope.
5. Using tools — reading data from a connected system or proposing an action, if the task requires it.
6. Human consent — a write operation covered by the write guard is held pending the user's decision.
7. Answer — the result is assembled, scanned and presented with citations, or with a note that evidence is missing.

What the user sees

• Citations and sources on answers grounded in knowledge.
• Attachments to the conversation, with content extraction for supported formats.
• Clarifying questions before an answer, when a key piece of information is missing.
• Information about tool limits, including web research.
• An action-consent card with the service, account and operation details.
• The ability to stop generation, rate an answer and ask a follow-up question.

The current interface has no "regenerate" button and no toolbar for flagging a correction by selecting a fragment of an answer — these are deliberately omitted elements that aren't in the product yet.

Status