LLM Chatbot custom visual

[LB_Team]

Hello everyone,

I’m working on a Power BI dashboard that uses a complex dataset (sourced from Tally) with multiple relationships, bridge tables, and advanced measures. I’d like to embed a chatbot directly into the Power BI report that can:

1. Understand the existing visuals, filters, and report context.

2. Answer natural-language questions

3. Leverage the underlying data model, not just relying on what’s displayed.

4. Be scalable and maintainable over time.
   

I have explored some options and one such is 

• LLM Chatbot custom visual by Creative Data
  Offers “Dataset” mode to understand report schema, but appears unstable with recent Power BI updates,many users report it fails to retrieve schema correctlyso suggest me some other methods which will adapt to my requirements and it should work robust to my data schema

[v-veshwara-msft]

Hi @LB_Team ,

Just checking in to see if you query is resolved and if any responses were helpful.
Otherwise, feel free to reach out for further assistance.

Thank you.

[v-veshwara-msft]

Hi @LB_Team 
We wanted to kindly follow up regarding your query. If you need any further assistance, please reach out.
Thank you.

[v-veshwara-msft]

Hi @LB_Team ,
Just wanted to check if the responses provided were helpful. If further assistance is needed, please reach out.
Thank you.

[v-veshwara-msft]

Hi @LB_Team ,

Thanks for reaching out to Microsoft Fabric Community.

The LLM Chatbot custom visual can be a useful way to bring natural language interaction directly into a Power BI report by allowing users to ask questions based on the underlying data model.

But as you already mentioned, the Dataset option in the visual is currently not functional due to recent updates in Power BI. You can try checking if it still works with earlier versions of Power BI Desktop.

In terms of alternatives, as mentioned by @DataNinja777  and @rohit1991 the built-in Q&A visual in Power BI can respond to natural-language questions using the model’s metadata, though it works best with well-structured synonyms and field names.

Copilot in Power BI offers a more advanced experience by understanding the underlying semantic model, including relationships, measures, and filters. It can assist in generating insights, creating visuals, and answering natural-language questions directly based on your dataset, which makes it a strong option for scalable and intelligent report interaction.

Overview of Copilot for Power BI - Power BI | Microsoft Learn

If you're looking for a more customizable setup, approaches using Azure OpenAI with Power BI via APIs or Power Virtual Agents can also be considered, depending on how much flexibility and integration is needed.

Here are some additional approaches and discussions that might help:
Re: ChatBot integration - Microsoft Fabric Community

Solved: Re: Chat widget integration in power bi dashboard - Microsoft Fabric Community

Please go through the links and see if any align with your requirements.

Hope this helps. Please reach out for further assistance.

Thank you.

[rohit1991]

Hi @LB_Team ,

For your needs, the most robust way to embed a chatbot in Power BI that understands visuals, filters, and the underlying data model is to use the Power BI Copilot feature (if available in your tenant) or embed a Power Apps chatbot (connected to Azure OpenAI or Power Virtual Agents) inside your report. The LLM Chatbot custom visual is unstable with complex schemas and recent updates, as you noted.

Direct solutions:

• Power BI Copilot: Native, context-aware, and directly understands report visuals and data model. Learn more.

• Power Apps Chatbot: Build a custom chatbot with Power Virtual Agents or Azure OpenAI, embed it in Power BI with the Power Apps visual, and connect it to your model via APIs or direct queries for true data context.

These options are more reliable and scalable than current 3rd-party chatbot visuals.

[DataNinja777]

Hi @LB_Team ,

I can see that dealing with an unstable custom visual for a critical function like a chatbot can certainly be frustrating, especially when your Power BI dashboard is built upon a complex dataset from Tally with intricate relationships and measures. T

To achieve a more reliable and scalable integration, you can explore several alternative methods that offer deeper integration with your data model and are more resilient to Power BI updates.

A great starting point is to maximize the use of Power BI's own native features, specifically the Q&A visual. This tool is designed to let users ask natural language questions and is inherently robust because it is a core part of the Power BI service. It directly accesses your entire underlying data model, including all its tables, relationships, and advanced measures, rather than just scraping visual-level data.

You can significantly enhance its understanding of your specific schema by configuring field synonyms, which teaches the Q&A engine your business's vocabulary, and by reviewing user questions to train the model over time. While not a conversational chatbot in the traditional sense, a well-configured Q&A visual provides a highly stable and maintainable way to enable natural language data exploration.

For a true conversational experience built on a low-code platform, Power Virtual Agents (PVA) offers an excellent balance of user-friendliness and power. The key to its robustness is its ability to trigger Power Automate flows. Instead of relying on a fragile connection to the report's front-end, you can design a bot in PVA that calls a flow whenever a user asks a question. This flow then uses the official Power BI connector to run a specific DAX query against your dataset via the stable REST API. The results are passed back to the bot, which then presents the answer to the user.

This architecture separates the chatbot's logic from the Power BI report, meaning it is not susceptible to breaking during Power BI user interface updates and can be maintained independently, providing the scalability you need.

For the highest degree of customization and ultimate resilience, the premier approach is to build a custom solution using the Azure Bot Service. This method completely decouples the chatbot from Power BI's front-end by hosting it as a separate web application. The bot leverages powerful services like Azure AI Language for sophisticated natural language understanding, allowing it to parse complex user queries. It then communicates securely with your dataset using the Power BI REST API to execute any DAX query imaginable, giving it the deepest possible understanding of your data model. Because the chatbot is simply embedded within the Power BI report as a web view (for instance, in an iframe), it is completely immune to changes in the Power BI service's interface. This pro-code option requires more development effort but provides a professional-grade, highly scalable, and exceptionally maintainable solution perfect for complex enterprise requirements. Given your needs, Power Virtual Agents likely provides the ideal blend of a robust API-driven backend and a manageable development experience, while the Azure Bot Service remains the definitive path for maximum control and long-term stability.

Best regards,