optimistic-gold•17mo ago
KB not indexing the whole website.
Hey, I am very new to VoiceFlow and tried to use a website as a Knowledge Base. The Website I used was: https://www.mfr.eu/services/
Below I will paste in what was indexed. When you go on the website you will see that it just took like 1/3 of the whole website. Anyone knows the reason?
And what was indexed:
Services - MFR - Automations - Chatbot Development Our Chatbot Development service crafts intelligent, custom chatbots designed to enrich the customer experience with responsive, 24/7 interaction. Tailored to your brand and support goals, our chatbots effectively address customer inquiries, leading to increased engagement and streamlined operations. Custom knowledge-base for precise responses Established pipeline tools for automatic lead capturing Custom walkthroughs for acquiring new clients AI Consulting Our AI Consulting service offers expert guidance to harness the transformative power of artificial intelligence for your business. Find out if your business can profit from the rapid AI evolution happening right now, which services or products would suit you the best or get advice on your existing AI strategy. Learn how your business can benefit from AI Without any obligation The initial consultation is completely free of charge https://www.mfr.eu/ https://www.mfr.eu/contact/ https://www.mfr.eu/ https://www.mfr.eu/services/ https://www.mfr.eu/about/ https://www.mfr.eu/contact/ https://www.mfr.eu/ https://www.mfr.eu/ https://www.mfr.eu/services/ https://www.mfr.eu/about/ https://www.mfr.eu/contact/ https://www.mfr.eu/contact/#booking https://www.mfr.eu/ https://www.mfr.eu/ https://www.mfr.eu/about/ https://www.mfr.eu/services/ https://www.mfr.eu/contact/ https://www.mfr.eu/impressum/ https://www.mfr.eu/privacy-policy/ https://www.mfr.eu/ https://www.mfr.eu/services/ https://www.mfr.eu/about/
2 Replies
The Knowledge Base (KB) in Voiceflow works by parsing and indexing the documents or websites you provide. However, there are a few reasons why it might not index the entire website.
Firstly, the KB parser reads the document or website and breaks it down into "chunks" of content using different techniques. If the website content is not structured in a way that the parser can understand, some parts might not be indexed.
Secondly, the KB parser uses an embedding model to convert each chunk into a vector representation of its meaning. This process costs money per token, and to manage costs, Voiceflow uses a
maxChunkSize parameter to moderate the size of the chunks. If a section of the website exceeds this size, it might not be indexed.
Lastly, the KB parser securely stores the chunks in a vector database. If there are any issues with storage, some chunks might not be indexed.
In your case, it seems like only a portion of the website https://www.mfr.eu/services/ was indexed. It could be due to the reasons mentioned above or other factors like the website's layout, content structure, or even server-side restrictions that prevent the entire site from being crawled and indexed.
To resolve this, you might want to check the structure and layout of your website to ensure it's compatible with Voiceflow's KB parser. If the issue persists, consider breaking down the website content into smaller sections or documents and upload them separately to the KB.
Remember, the more documents you upload, the better the AI agent can use them to answer questions. However, there's a 10mb file limit per upload, so make sure your documents are within this limit.
For more information on how the Knowledge Base works, you can refer to the Voiceflow documentation here and here.optimistic-goldOP•17mo ago
I am pretty sure the website is not too long. I was using Botpress previously and wanted to try out VoiceFlow. In Botpress the whole website was indexed correctly. Furthermore there is really not much content on my website