Enter [Development Space], click [Create Bot], and select the Bot type: Knowledge Q&A and AI Assistant.

  • Knowledge Q&A: With a "short memory", it is suitable for scenarios that focus on "one question and one answer", such as translation, retrieval, judgment, etc.

  • AI Assistant: In addition to "short memory", it also has a "long memory" function, that is, it can continue the conversation based on a long conversation history, which is suitable for chatting and other scenarios.

  • Input the Bot name and logo;

  • Bot welcome message: when the user enters the Bot dialogue interface, the first sentence that the Bot sends to the user;

  • Bot identity prompt: set the role of the Bot and tell the Bot what identity it needs to handle problems for you. The system initialization of the identity setting is very important for the Bot, and it will be used to shape the Bot's identity, capabilities, boundaries and emotions, etc. It is recommended that the identity prompt text be more precise, organized, clear and logical. You don't need to pay too much attention to the amount of content, the higher the quality of the identity prompt text, the higher the quality of the AI response. Different separators can be used to distinguish different types of prompt content, so that the AI can clearly distinguish. Provide correct examples as much as possible to help AI output content correctly.

  • Model Selection: Select the model and its version to use. Different model versions consume different points, please refer to "Service Pricing" for details;

  • Temperature: the creativity of Bot in answering questions. A lower value means that the AI output tends to be more stable and rigorous, and a higher value means that the AI output tends to be more unrestrained and creative.

  • Doc Correlation: the correlation value between the content of the document retrieved from the Bot document and the user's question based on the user's question. Bot will only recall the content of documents whose relevance is higher than this value as corpus, so the higher the value, the more accurate the matching knowledge, but the amount of knowledge may be less;

  • Matching knowledge: the upper limit of the number of document content retrieved from the Bot document based on the user's question. The Bot will only recall up to that amount of document content;

  • Response to irrelevant questions: when the Bot cannot find relevant document content in the Bot document, whether the Bot will answer the user's question according to the model's understanding of the user's question;

  • Display knowledge sources in responses: in the user dialogue interface, the answer given by the Bot, whether to display the content source of the Bot document of the answer.

  • Long-term memory: If it is enabled, when the Bot answers, it will combine the longer chat history as the context to answer;

  • Short-term memory: When Bot answers, it will combine the recent rounds of chat records as context to answer;

  • Suggested Questions: If enabled, the BOT will provide the user with 3 suggested questions after each answer. This feature consumes credits.

After the selection is complete, click [Create] to complete the Bot creation.

Last updated