Create a Gen AI Bot

this comprehensive user manual guides you through the process of developing a gen ai chatbot using the bot designer it covers the architectural overview, step by step construction, common use cases, and essential best practices to ensure optimal performance 1\ introduction to the gen ai chatbot the gen ai chatbot is engineered with a sophisticated two tiered architecture delegator (bot settings)/ supervisor this is the bot's central intelligence it analyzes user input to identify the correct objective (user intent) and directs the query to the appropriate "goal " the delegator can also respond directly if it possesses sufficient context individual goals (outcomes) these are specialized agents that handle specific user queries they effectively replace traditional natural language processing (nlp) intents by allowing the delegator to dynamically determine the user's intention based on conversational context currently, the architecture supports a single level of goals, meaning sub goals are not permitted when you select either the delegator or a specific goal in the bot designer, a side panel appears, providing access to three crucial configuration tabs objectives , settings , and tools 2\ constructing your gen ai chatbot building your gen ai chatbot involves a systematic configuration process of the delegator and its associated goals 2 1 defining core bot instructions begin by outlining the fundamental instructions for your bot within the delegator's objectives and rules fields it's important to understand that while these are presented as separate fields, their content is combined and fed to the gen ai bot as a single, unified input this input must be comprehensive, including all necessary bot instructions and any references to tools, to ensure accurate and relevant execution 2 2 adding and naming goals next, you will add individual goals to your bot this is done by selecting the "add subgoal" option within the delegator's interface for each new goal you create, navigate to its settings tab and provide a short, crisp name this naming convention is crucial as it helps the delegator quickly and accurately identify the correct objective during a live conversation 2 3 configuring goal instructions within each goal, access the objectives tab to outline specific instructions pertinent to that particular outcome these instructions guide the goal on how to process user queries and interact with tools 2 4 enabling and configuring ai agent tools to empower your goals to interact with external functions, toggle the ai agent switch to on within the goal's settings tab once enabled, you can configure external tools in the tools tab the bot designer supports two primary types of tools exomind and apis a single goal can be configured to utilize multiple tools key tool configuration details name provide a clear and concise name for your tool crucially, tool names cannot contain spaces the ai agent uses this name to decide when to invoke the tool description write an accurate and detailed description of what the tool does this description is vital for the ai agent to understand the tool's functionality and determine its relevance to a user query for an api tool, you will need to specify additional details url the endpoint for your api call request type currently, only json based rest api calls are supported arguments the parameters required by your api headers any necessary http headers for the api request response variable the variable that will store the api's response 2 5 advanced tool use force action and make mandatory force action located in the goal's settings , the force action toggle is used for scenarios where an external data call is absolutely essential when enabled, it compels the goal to call its configured tool at least once before generating a response for the user make mandatory also found in the goal's settings , the make mandatory toggle is used if the bot must fulfill this specific goal during every user session 3\ common use cases the gen ai chatbot architecture is ideally suited for complex conversational flows that demand context aware delegation and integration with external functionalities customer support & information retrieval configure goals to delegate queries to an api to fetch real time information, such as order status, account balances, or product availability alternatively, use an exomind tool to retrieve information from a knowledge base forced action scenarios for situations where an external data call is non negotiable (e g , checking inventory before confirming an item's stock), enable force action for the relevant goal to ensure the necessary api call is executed 4\ best practices for optimal performance adhering to these best practices will significantly enhance your gen ai chatbot's performance, accuracy, and maintainability instruction clarity feed all necessary instructions, including explicit references to tools, into the delegator's objectives and rules fields the more comprehensive and clear your instructions, the better the bot will understand its role and execute tasks concise goal naming use short and crisp names for your goals this makes it easier for the delegator to quickly and accurately identify the correct objective for any given user query precise tool configuration the tool's name and a detailed description are critical for the ai agent's decision making process ensure these are accurate and unambiguous remember that tool names cannot contain spaces leverage exomind for accuracy when retrieving information from a knowledge base, prioritize using exomind a key advantage of exomind is that prior results are not included in the chat history shared with the large language model (llm) this leads to a smaller prompt size for the llm, which generally translates to increased accuracy and more efficient response generation advanced api handling with functions for complex api interactions, such as those involving nested json request bodies or the need to parse nested key value pairs from a response, it is a best practice to use in house functions to manage the api logic avoid attempting to configure these complexities directly within the bot designer's api fields, as this can lead to errors and reduced maintainability