Conversational AI

Co.chat allows users to chat with the model, without worrying about prompt engineering. This endpoint is used to create conversational experiences with AI. This guide provides a starting point for using the endpoint.

You can use it by joining the Coral waitlist here.

You can generate the next turn of a conversation using this endpoint. The new turn takes into account the previous messages and the specified parameters.

The endpoint has a number of settings you can use to control the kind of output it generates:

• message: (String, Required) The message that the user is sending to the chat endpoint.
• model: (String) The name of the model to generate with.
• preamble_override: (String) The text to replace the default preamble for each response
• chat_history: (Object) This JSON object contains the previous messages in a conversation, formatted as an array of objects.
• user_name: (String) The name of the user.
• stream: (Boolean) Whether to stream the response.
• temperature: (Number) The degree of randomness of the response.
• max_tokens: (Integer) The maximum amount of tokens to generate. This parameter is optional, and not specified by default.
• return_preamble: (Boolean) Whether to return the preamble.
• return_prompt: (Boolean) Whether to return the prompt.

For more details refer to the API Reference.

Starting a conversation