Custom File Upload for Answer Generation to Power Virtual Agents

Summarized Video on Chatbot Answers Generation from Uploaded Documents

For those wanting to leverage custom chatbots, AI and Machine Learning capabilities have taken a leap forward with a feature now in preview that allows for the upload of documents for the chatbot to generate its responses from. This YouTube video, presented by "Daniel Christian [MVP]", focuses on demonstrating this feature.

The narrator shows how one can upload documents which the chatbot can use to generate answers using generative AI. When the bot can't find a relevant topic in its pre-defined list, it will draw an answer from the uploaded documents' contents.

1. Introduction 00:00
2. Announcement 01:20
3. Building The Chatbot 02:15
4. Upload the documents 03:03
5. Enable use of generative answers 04:14
6. Testing the Chatbot 05:03
7. Content sharing awareness 06:54
8. Where are the documents saved? 09:53
9. Build the canvas app 11:01
10. View Bot Dataverse tables 16:20
11. Conclusion 18:10

The feature uses advanced AI models to find the best answer within the uploaded documents and presents it in a conversational style. The data from the uploaded documents will be safely kept in Dataverse and the number of uploads is restrained by the available storage in your Dataverse environment.

However, it's essential to note that this feature is currently just a preview and not meant for production use. And, the feature might have restricted functionality being in its early phase. Users should also be aware that the content of the files they upload will be accessible to all bot users, regardless of file permissions or access controls.

To use this feature, various document types can be uploaded like Word documents, Excel spreadsheets, PowerPoint documents, PDF documents, Text, HTML, CSV, XML, OpenDocument files, EPUB, Rich Text Format documents, Apple iWork documents, JSON files, YAML files, and LaTeX files. Certain types, like images, video, executable files, and audio, are not supported at this time.

Once the documents are uploaded, it takes a few minutes to prepare the document content to be used by the bot. The documents, even with the same name, don’t overwrite existing documents, instead, new copies are created. Therefore, when updating a file, it’s advised to add a version number to the file name, distinguishing it from previous versions.

The documents used as the data source for the bot will be used as a default for all generative answer nodes. If a specific node should not use the uploaded documents, it can be done by inserting specific lines of code presented in the video. Also, the uploaded documents become part of the bot solution, which are included when exporting or importing a bot solution.

Significance of AI + Machine Learning in Chatbots

AI and Machine Learning abilities have revolutionized the way chatbots work, allowing them to be more versatile, intuitive, and more conversational. By sourcing their responses from uploaded documents, bots can now provide much more precise and helpful responses. These advancements are paving the way for richer, more natural conversations between bots and users, elevating user experiences and making chatbots an even more valuable resource for organizations.