Skip to main content

Chat

Chat is a place to evaluate the answer quality that your application can expect from similar API requests. It is a testing tool, not a general personal assistant, an everyday AI workspace or a production integration. It keeps only a short recent context, so use New topic when moving to another independent test.

Writing a useful request​

Describe one concrete result and include what the answer must use:

  • the goal and audience;
  • the language and required format;
  • source text, facts or links;
  • constraints and anything that must not be invented;
  • one main deliverable per request.

Omev applies an individual quality policy to the account automatically. The more you use Omev AI, the better it understands recurring needs and classifies requests. You do not need to configure that policy yourself.

What the cabinet shows​

While the answer is prepared, the side panel explains how Omev understood the request and how confident that reading is. If the interpretation is wrong, add the missing detail and send the request again.

When live web sources are used, they appear below the answer with links. Review those sources before publishing the result or using it for a high-impact decision.

Result formats and feedback​

Depending on the request, the result can be formatted text, plain text, structured JSON, or a document together with structured data. Available actions let you copy the result or download .md and .json files for evaluation. These actions do not turn Chat into a production work tool.

Use the positive or negative rating under an answer. For a negative rating, briefly state what was wrong in the meaning, facts or format. This feedback is more useful than resending an unchanged request many times.

History​

Open History to return to previous conversations. A conversation ends when you choose New topic or after a long inactive period. Reloading the page does not erase saved history.

Test the same request as code​

Move to Playground when you need to verify the technical compatibility of the exact JSON body, parameters and response shape that your application will use, then copy the integration code. Playground is not a production queue or an alternative interface for ongoing AI work. See Getting started for the complete API path.