Skip to main content
In the XpertAI platform, Dimension Member Retrieval is a key capability module in the ChatBI agent orchestration system. By dynamically synchronizing and vectorizing dimension member information in semantic models, it provides agents with more precise context awareness and intent recognition support, greatly improving the accuracy and user experience of conversational data analysis. In short, dimension member retrieval enables agents to “know what possible dimension values exist” and “understand the semantic meanings of these dimension members in different business contexts.”

​
I. Synchronization and Vectorization of Dimension Member Information

To enable dimension member retrieval functionality, you need to complete the synchronization and vectorization of dimension data in semantic models first. The entire process is as follows:

​
1. Dimension Member Synchronization (Sync)

After creating or updating a semantic model, manually trigger the synchronization task for dimension members. The process includes:
  • Refreshing the member list of dimensions.
  • Uploading and embedding dimension members for subsequent processing and retrieval.

​
2. Vectorization Processing (Embedding)

After synchronization is complete, the platform will perform semantic vectorization processing on dimension members. The steps include:
  • Using built-in or user-defined vector models (such as OpenAI Embedding, Tongyi embedding, etc.)
  • Converting each dimension member (and its aliases, business labels) into semantic vectors
  • Storing vectorization results in the platform’s embedded vector database (supports (in development) FAISS / Qdrant / Milvus, etc.)

​
II. How to Use in ChatBI Toolset

After vectorization is complete, this dimension member information will be integrated into the ChatBI agent’s toolchain through the Dimension Member Retriever tool, with the following specific functions:

​
Agent Call Flow:

  1. Intent Recognition Phase: The agent determines whether the user’s query contains vague dimension descriptions (such as “East China”, “xx customer”, “holidays”)
  2. Retrieval Call Phase: Through Dimension Member Retriever, perform semantic vector comparison to find the closest matching items from the dimension member library
  3. Result Lookup Phase: Use matched dimension values as filter conditions to participate in query construction and chart generation

​
Application Examples:

  • User asks: “How are sales in East China this month?”
    • Agent matches “East China” through Retriever → region = 'East China'
  • User asks: “Which products sold well during holidays?”
    • Matches “holidays” dimension members → Automatically incorporates the corresponding time periods for holidays in the analysis
🧠 ChatBI performs fuzzy recognition and contextual reasoning based on dimension business labels, pinyin, English aliases, etc.

​
III. Business Value and Advantages

FunctionBusiness Advantage
Dynamic Synchronization of Dimension MembersEnsures real-time updates of dimension member information, reducing missed queries
Vectorized RecognitionImproves natural language recognition accuracy, supports fuzzy matching
Multi-language UnderstandingCovers international scenarios, recognizes dimension names in multiple languages
Business Semantic AwarenessIntelligently understands business aliases, abbreviations, and classification labels
No Rule Maintenance RequiredCompared to traditional keyword matching solutions, no manual maintenance of mapping tables needed

​
IV. Usage Notes

  1. Data Source Access Permissions: Ensure that data tables from which dimension members originate have complete read permissions;
  2. Data Volume and Performance: For million-level dimension member data, it is recommended to configure vector extraction frequency and index partitioning as needed;
  3. Language Consistency: If there are scenarios with mixed Chinese and English, please set member multi-language labels or alias fields in the semantic model;
  4. Regular Update Mechanism: For business dimensions (such as customers, products), it is recommended to set up regular synchronization schedules (such as daily or hourly);
  5. Privacy Compliance: If dimensions contain sensitive information such as customers and users, proper desensitization and permission isolation controls must be implemented;

​
V. Future Plans

We plan to include the following in future versions:
  • Support for user-defined dimension member semantic labels
  • Introduction of dimension member weights and popularity metrics to enhance matching accuracy
  • Provide a dimension member management backend supporting manual editing and review