Voiceflow is unable to extract information from dashboard
Whenever I am adding the dashboard URL to my chatbots (Voiceflow) knowledge base, it is unable to read or extract information from that particular URL. However, it can read the same information from Google Sheets.
2 Replies
Hi there!
Voiceflow’s URL document-type can only fetch the raw HTML that’s returned when it performs a normal, unauthenticated GET request to the link you provide.
If the page you’re pointing to … • sits behind a login or token-based security layer, OR
• is rendered dynamically in the browser with JavaScript (common for analytics dashboards such as Looker, Power BI, Tableau, etc.)
… then the Voiceflow crawler receives almost no text to index, so the document ends up empty in the Knowledge Base. A published Google Sheet works because it’s a publicly accessible HTML page that returns static text right away, so the crawler can read and chunk it without issues. How to work around it 1️⃣ Export the dashboard data to a static format (CSV, TSV, TXT, PDF) and upload that file to the Knowledge Base instead.
2️⃣ Publish the dashboard (or a read-only view) to a public URL that doesn’t require sign-in and serves static HTML.
3️⃣ If the dashboard has an API, you can call it in a Function step, format the response, and add the result to the KB via the “Docs / upload” endpoint. Reference: “Best Practices for the Knowledge Base – Web documents (HTML) are generally well-parsed and are the recommended format for URL uploads.” (docs.voiceflow.com) Hope this clarifies why the dashboard link isn’t being ingested and gives you some options to move forward!
If the page you’re pointing to … • sits behind a login or token-based security layer, OR
• is rendered dynamically in the browser with JavaScript (common for analytics dashboards such as Looker, Power BI, Tableau, etc.)
… then the Voiceflow crawler receives almost no text to index, so the document ends up empty in the Knowledge Base. A published Google Sheet works because it’s a publicly accessible HTML page that returns static text right away, so the crawler can read and chunk it without issues. How to work around it 1️⃣ Export the dashboard data to a static format (CSV, TSV, TXT, PDF) and upload that file to the Knowledge Base instead.
2️⃣ Publish the dashboard (or a read-only view) to a public URL that doesn’t require sign-in and serves static HTML.
3️⃣ If the dashboard has an API, you can call it in a Function step, format the response, and add the result to the KB via the “Docs / upload” endpoint. Reference: “Best Practices for the Knowledge Base – Web documents (HTML) are generally well-parsed and are the recommended format for URL uploads.” (docs.voiceflow.com) Hope this clarifies why the dashboard link isn’t being ingested and gives you some options to move forward!
Why would you try to add the dashboard url to your agent's KB?