HENGSHI ChatBI User Manual

Product Overview

HENGSHI ChatBI is an intelligent data analysis tool integrated with AI technology, designed to provide business professionals with an intuitive and efficient data interaction experience. Through natural language processing technology, users can directly interact with data, quickly obtain the required information, and thereby provide strong support for business decision-making. Additionally, HENGSHI ChatBI supports private deployment, ensuring the security and privacy of enterprise data.

Installation and Configuration

Prerequisites

Before you start using HENGSHI ChatBI, please ensure the following steps are completed:

1. Installation and Startup: Follow the Installation and Startup Guide to complete the installation of the HENGSHI service.
2. AI Assistant Deployment: Follow the AI Assistant Deployment Documentation to complete the installation and deployment of related services.

Configure Large Model

After the HENGSHI service starts, go to the "Feature Configuration" page in System Settings to configure the relevant information for the AI Copilot, including the address and key of the large model.

Don't know how to configure? Please refer to FAQ.

User Guide

Enhancing the Understanding Capability of Large Models

To ensure that ChatBI can accurately comprehend your business requirements, it is recommended to configure the following:

1. Enhance Understanding of Company Business, Industry Terminology, and Private Domain Knowledge

In the AI Assistant Console under System Settings, use natural language to describe your business scenarios and terminology in the UserSystem Prompt. Ensure that the Enable Model Inference Intent in the general model configuration is turned on.

For example, if you need to prohibit answering certain types of questions, you can specify in the prompt, "Do not answer questions related to revenue."

2. Enhancing Data Understanding

• Dataset Naming: Ensure dataset names are concise and clearly reflect their purpose.
• Knowledge Management: Provide detailed descriptions of dataset purposes, implicit rules (e.g., filter conditions), synonyms, and business-specific terms corresponding to fields and metrics in Knowledge Management.
• Field Management: Ensure field names are concise and descriptive, avoiding special characters. Provide detailed explanations of field purposes in Field Descriptions, such as "Default field for timeline."
• Metric Management: Ensure atomic metric names are concise and descriptive, avoiding special characters. Provide detailed explanations of metric purposes in Atomic Metric Descriptions.
• Field Hiding: For fields not involved in Q&A, it is recommended to hide them to reduce the number of tokens sent to large models, improve response speed, and lower costs.
• Distinguishing Fields and Metrics: Ensure field names and metric names are not similar to avoid confusion. Fields not required for answering questions should be hidden, and unnecessary metrics should be deleted.
• Data Vectorization: Publishing an app will trigger the intelligent data vectorization task for the dataset. You can also manually trigger the "Intelligent Data Vectorization" task. This task deduplicates field values and vectorizes the dataset to improve filtering accuracy.
• Intelligent Learning: It is recommended to trigger the "Intelligent Learning" task to convert general examples into dataset-specific examples. After execution, manually review the learning results and perform additions, deletions, or modifications to enhance the assistant's capabilities.

3. Enhancing Understanding of Complex Calculations

For complex aggregate calculations, it is recommended to define them as Metrics to reduce the complexity of data retrieval in the model and avoid misunderstandings of private domain knowledge by large models.

For example, the calculation method for ROI in advertising companies differs from that in manufacturing. However, large models cannot automatically recognize these differences. Therefore, it is recommended that you create a metric and provide a detailed description of its meaning to ensure that the large model does not invent its own calculation formula during data retrieval.

Usage Scenarios

1. Go to Analyze

Go to Analyze is an enhanced feature of HENGSHI SENSE Chart. The system integrates the Metrics Analysis Function with the published App, enabling the published Chart to have secondary analysis capabilities.

Quick Start

1. Log in to the System: Open your browser, navigate to the HENGSHI ChatBI login page, and enter your account and password.
2. Configure AI Assistant: Go to "System Settings" > AI Assistant Configuration, and input the address and key of the large model. (Requires system administrator role)
3. Create an App: On the "App Creation" page, click Create New App to create a blank app.
4. Create a Dataset: On the "Dataset" page, click Create New Dataset to upload your data or connect to your data via Data Connection.
5. Create a Dashboard: In the app, create a dashboard, add charts, and select the dataset you just created as the data source.
6. Publish the App: After completing chart creation, click Publish App to publish the app to the App Marketplace. Check the Go Analyze feature during publishing.
7. Go Analyze: In the App Marketplace, click the published app to enter the app details page. Click the Go Analyze button in the upper-right corner of the chart to access the secondary analysis feature page.
8. Start a Conversation: In the ChatBI interface, input your question, such as "Show last month's sales."
9. View Analysis Results: The system will generate charts or tables, allowing you to interact and perform further analysis directly on the interface.

2. Conversing in the Dashboard

By leveraging the Global JS Functionality and dashboard Control Events, you can embed ChatBI into the dashboard, enabling users to interact directly with the data, gain insights, and perform secondary analysis.

Quick Start

1. Enable SDK: Log in to HENGSHI ChatBI, navigate to "System Settings" > Global JS Functionality, and refer to Integrating Copilot into Dashboards within the HENGSHI System to configure the code properly.
2. Follow the same steps as in Quick Start for Analysis, which involves creating an App and a Dashboard.
3. Add a Button: Add a button to the Dashboard and set the button's Control Events. Refer to Integrating Copilot into Dashboards within the HENGSHI System.
4. Click the Button: Click the button on the Dashboard to open the ChatBI window and perform conversational analysis.

3. Easily Integrate ChatBI

HENGSHI SENSE offers multiple integration methods, allowing you to choose the one that best suits your needs:

• iframe Integration: Use iframe to integrate ChatBI into your existing system, achieving seamless connectivity with the HENGSHI SENSE BI PaaS platform.
• JS SDK Integration: Achieve more precise control, such as custom UI, through the JS SDK. The SDK provides a wealth of configuration options to meet personalized needs.
• ReactJs SDK Integration: Easily integrate ChatBI into React projects using the ReactJs SDK, without worrying about component reuse and style conflicts.
• API Integration: Integrate ChatBI into backend systems through the API to implement more complex business logic.
• Feishu, DingTalk, and WeCom Integration: Use the backend API to integrate ChatBI into your Feishu, enabling customized ChatBI business logic.

FAQ

How to Troubleshoot Model Connection Failures?

There are various reasons for connection failures. It is recommended to troubleshoot step by step as follows:

1. Check the Request URL Ensure the model URL is correct. Different vendors provide different model URLs. Please refer to the documentation provided by the vendor you purchased from.

   We can provide some initial guidance:

   • Model URLs from various vendors usually end with <host>/chat/completions rather than just the domain name, for example, https://api.openai.com/v1/chat/completions.
   • If your model vendor is Azure OpenAI, the model URL is structured as https://<your-tenant>.openai.azure.com/openai/deployments/<your-model>/chat/completions. Here, <your-tenant> is your tenant name, and <your-model> is your model name, which you need to check on the Azure OpenAI platform. For more detailed steps, please refer to Connecting to Azure OpenAI.
   • If your model vendor is Tongyi Qianwen, there are two types of model URLs: one compatible with the OpenAI format, https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions, and another specific to Tongyi Qianwen, https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation. When using the OpenAI-compatible format, please select OpenAI or OpenAI-API-compatible as the provider in the HENGSHI Intelligent Query Assistant model configuration.
   • If your model is privately deployed, ensure the model URL is correct, the model service is running, and the model provides an HTTP service with an interface format compatible with the OpenAI API.

2. Check the Key

   • Most large model interfaces provided by vendors require a key for access. Ensure the key you provided is correct and has permission to access the model.
   • If your company has deployed its own model, a key might not be required. Please confirm this with your company's developers or engineering team.

3. Check the Model Name

   • Most vendors offer multiple models. Choose the appropriate model based on your needs and ensure the model name you provided is correct and has the necessary access permissions.
   • If your company has deployed its own model, a model name might not be required. Please confirm this with your company's developers or engineering team.

How to troubleshoot errors when querying data?

1. Is the vector database installed? If not, please follow the AI Assistant Deployment Documentation to complete the installation and deployment of related services.
2. Can the model connect? Follow the troubleshooting steps in the previous question to check whether the model can connect.

How to Fill in the Vector Database Address?

Follow the AI Assistant Deployment Documentation to complete the installation and deployment of related services. Manual input is not required.

Does it support other vector models?

Currently, it is not supported. If you have any requirements, please contact the support engineer.