AI Configuration
Momen AI: Based on LLMs, leveraging RAG and tool invocation, you can create customized business AI apps and agents.
Introduction
Optimize your business efficiency with Momen's AI capabilities. Learn about custom AI interfaces, knowledge base integration, and output configurations. Explore the potential of Momen AI in various scenarios in this tutorial.\
Function Description
Momen AI is a robust feature integrated within the Momen editor, designed to facilitate custom AI request setups. Its streamlined interface demystifies AI configurations, enabling you to create personalized AI applications without coding.
Core Features
Supports a variety of large models, including "GPT 3.5 turbo" and "GPT 4o". Momen will integrate more models in the future.
Accommodates multiple input types, such as text and images.
Connects to various knowledge base sources, including the Momen database, third-party APIs, and local files.
Offers a selection of tools for invocation, including behavior flows, APIs, and other pre-configured AI requests.
Provides outputs in JSON format and supports streaming outputs.
Use Cases
AI applications have a wide range of use cases. Here are some common scenarios:
Intelligent Customer Service (AI Q&A):
Insurance, law firms, tools, e-commerce AI Q&A assistants/intelligent customer service.
Provide AI with specialized knowledge within an industry, such as uploading Momen's docs for AI to answer questions related to Momen's operations.
Marketing (AI Content Creation):
Sales/consultation script generation based on user information and sales/consultation materials.
Copywriting for platforms like Ins, X, and Tiktok and scriptwriting for videos, movies, etc.
Function Entry
Click on the Momen AI button in the editor's top left corner to enter the AI configuration interface. Click "Create" to add a new AI request.
Configuration Instructions
The left side is the configuration interface, the right side is the debugging interface.
Defining the Request Name
AI Request Name: The default setting is Ai-Config-1. You can customize the name based on your business requirements, making it easier to identify during invocation.
General Configuration
Large Model: Select the version of the large model to complete your AI tasks. The version you choose can influence the results.
More sophisticated models will use more tokens during.
Currently, only GPT 4o supports processing images.
AI Personality: This refers to 'temperature' in configuration, which ranges from 0-1. A lower value yields more rational and accurate results, and a higher value yields more creative responses.
Maximum Rounds: This is the maximum number of interactions you can have with the AI during debugging. Sending a command to the AI and receiving a response without invoking any tools counts as one round. If you send a command and the AI calls an API tool before responding, that counts as two rounds.
Maximum Token per Round: This is the maximum number of words that can be sent to the AI and received from the AI in each round.
Important Note: When multi-turn conversations are enabled, each new round includes the context from previous rounds. This will result in an increasing number of words per round, consuming more tokens. Therefore, you need to set a word limit per round based on your business needs to avoid exceeding the model's processing capacity while efficiently using tokens.
Input Configuration
Input Parameter Configuration: Set up the single or multiple parameters required when making an AI request. These parameters are typically linked directly to [Predefined Prompts] for usage, and may also be used as filtering conditions for the [Knowledge Base-Database].
Preset Prompts
Prompts: Prompts serve to inform AI of its role, provide reference material, and specify task constraints.
Advanced Prompts Guide
It is recommended to read and understand the Prompt Engineering Guide to gain insight into prompt specifications and techniques for optimal results. While high-quality prompts may consume more App Tokens and waiting time, they often yield better results.
Prompt Source Options
Direct Input: Enter prompt words directly in task roles, messages, and constraints. Note that this configuration will make the AI only execute fixed tasks.
AI Parameter Reference: Click the "+" sign next to the input box to reference parameters that are passed when the AI is invoked. This setup is more flexible than direct input, allowing users to define their own task roles, details, and constraints, enabling the request to handle a wider array of tasks.
Direct Input + AI Parameter Reference: If your business involves assisting users in writing content, where task roles or constraints remain fixed and only task messages need user input, you can directly input fixed roles and constraint content in the AI interface. Task details can then be obtained using AI parameters.
Knowledge Base: Retrieve data from the Momen database, third-party APIs, or files uploaded locally.
Momen AI Prompt Content
To enhance prompt quality, Momen AI has structured the "preset prompts." New users can follow the structure below or reference input parameters to obtain relatively high-quality results.
Task Roles:
Define what role AI should play, facilitating better understanding of tasks/questions. For example, "An English teacher helping students master English grammar," "An insurance claims specialist," or "An influencer on Ins"
Task Details:
Define how AI should perform, whether answering questions, writing copies, or extracting information. Specify whether to integrate other context or knowledge base references when executing tasks.
Task Constraints:
Specify constraints AI should follow during task execution, such as what to do, what not to do, word count limitations, keywords to include, required style, precision level, and generation of rich text, among others.
Example:
Suppose the task is to create an AI assistant for Momen AI development. Users can ask questions related to Momen's usage. The AI assistant relies on the provided knowledge base documentation. If the question is not in the knowledge base, the response should be, "Sorry, I cannot answer your question. Please contact Momen experts for assistance."
Task Roles: "Assisting users in correctly using a no-code tool called Momen based on knowledge base documentation."
Task Messages Details: "Your task is to answer user questions based on the provided knowledge base content. The response must include the knowledge base original text and your summary based on the question. Format: [Knowledge base original text:, Summary:]"
Task Constraints: "Only answer user questions based on the provided knowledge base content. If there are no relevant answers in the knowledge base, respond with: 'Sorry, I cannot answer your question. Please contact Momen experts for assistance.'"
Knowledge Base
Function
Results retrieved from the knowledge base along with preset prompts will be sent to AI.
Expand Knowledge Domain: External knowledge bases provide additional domain-specific knowledge to enhance the model's ability to answer a broader range of questions.
Provide Real-time Information: External knowledge bases can include the latest data, aiding AI in understanding current events and trends.
More Uses: Improve answer accuracy, enhance contextual understanding, handle rare domain issues, language translation, cross-cultural understanding, and compensate for knowledge gaps.
Example:
Case 1: Create an AI assistant for Momen AI development. Users can ask questions related to the development and usage of Momen. To ensure accurate responses, AI needs access to Momen's official documentation.
Case 2: Develop an AI assistant for internal company finance processes. Colleagues can pre-ask the AI assistant questions related to financial reimbursements. To ensure precise responses, AI needs access to relevant documentation regarding company financial reimbursements.
Knowledge Base Data Source
Momen AI supports multiple data sources, including database, APIs and local files.
Database:
Select single or multiple data tables, fields, related tables/fields stored in Momen. You can also predefine data limits, filtering conditions, sorting, and deduplication operations on selected tables.
Important Note: The data given to the AI will consume tokens. To manage this, it's best to limit the amount and size of data entries. Set appropriate data limits based on your needs to avoid overloading the model and use tokens efficiently.
Local Files:
Upload your local files as AI's knowledge.
Supported formats: Word, PDF, txt, markdown.
Total file size should not exceed 50MB.
Third-Party API
Momen also supports fetching external data via API. You can choose the configured APIs as your data source.
Please check our step-by-step guide about how to configure APIs in Momen: Notion, Airtable, Google Sheet
Tools
Why does an AI agent need to Invoke Tools
In an AI Agent, large language models serve as the agent's brain, tasked with logical thinking and task planning. Yet, these models can't directly interact with the world; they require various tools like API requests or intermediary software to facilitate external interactions, similar to human limbs. These tools allow the AI to perform specific actions such as data retrieval, image generation, or other environmental interactions.
Types of Tools
In Momen AI, the AI can invoke pre-configured Actionflows, APIs, and other AI requests within Momen, or open "Obtain info" to let the AI seek additional context from the user if necessary.
Invoking Actionflows
Preparation: Familiarize yourself with [Actionflow Fundamentals], complete the Actionflow setup, save, and publish it. After choosing a configured Actionflow, click the add description button to the right of the flow to describe both the actionflow itself and the input/output parameters, helping the AI understand the flow's purpose and the data involved.
Invoking APIs
Preparation: Learn about [API Configuration], complete API setup and testing, save, and publish the API. After selecting a configured API, click the add description button on the right side of the API to detail what parameters to send and the expected API returns.
Invoking AI
Preparation: Other AI requests have been set up and released in Momen. Similarly to the above tools, after choosing another configured AI, add a description, informing the current AI about the tool's function and the input/output data involved.
Output Configuration
Default: If your task does not specify an output format, the AI will automatically determine the most appropriate output. For instance, it may produce a number when calculating data, or a substantial text block when writing an article.
Custom: As depicted below, outputs can be formatted as "JSON" objects. This format allows for the customization of the JSON structure and its field names. Each field can include a description that clarifies its purpose to the AI.
Streaming Output: AI models display results progressively as they are generated, reducing latency and thereby improving user experience.
Important Note: Streaming output is intended solely for visual display on pages and cannot be integrated into project data flows. It can be used as standard text for further applications, like database storage, only after a successful AI invocation.
Debugging and Preview
Once you've finalized the AI settings, input parameters can be entered on the right side of the interface for debugging and previewing the AI request.
The section below details the entire operation process of the AI. Upon successful execution, you can review the input parameters you provided, how the tools were utilized, and whether the outcomes align with your expectations. If the setup includes multiple dialogue rounds, you can input new parameters here to extend the conversation.
Request AI
App Token Consumption
About Momen
Momen is a no-code web app builder, allows users to build fully customizable web apps, marketplaces, Social Networks, AI Apps, Enterprise SaaS, and much more. You can iterate and refine your projects in real-time, ensuring a seamless creation process. Meanwhile, Momen offers powerful API integration capabilities, allowing you to connect your projects to any service you need. With Momen, you can bring your ideas to life and build remarkable digital solutions and get your web app products to market faster than ever before.
Last updated
