Using multiple Knowledgebases
Hey there,
we are setting up a chatbot, which should respond to customer questions to our certification system.
The issue is, that we have different versions of the system (2018 and 2023). And depending on the version, the correct response could differ.
So my plan was to get the users question and let him pick the correct version via click fields in the chat after that. With the question and the given version I would go to different knowledge bases, one for each version.
The issue is so far I couldnt find an option to maintain different knowledgebases and keep them seperate as the source of information are quite similar, but still different.
Is that somehow possible? Am I too concerned about the AI mixing the different versions up and I can just import both Versions into the same knowledgebase?
Thanks alot!
2 Replies
You can definitely keep the two sets of docs apart – you just do it with tagging rather than with two physical Knowledge Bases.
• One KB per project
Each Voiceflow project (agent) has a single Knowledge Base. You can’t create “KB-2018” and “KB-2023” side-by-side inside the same project. • How to keep the content separate
1. Upload the 2018 files → add a KB tag like 2018
2. Upload the 2023 files → add a KB tag like 2023 • Route the user the right way
1. In the conversation ask “Which version are you using? – 2018 | 2023”.
2. Save their choice to a variable (e.g. {version}).
3. Use a KB Search step (or the KB Query API) and pass a filter such as: Only chunks that carry the selected tag will be returned, so answers never mix. (If you rely on the Agent step’s automatic KB lookup you can’t pass a filter, so just toggle the KB off and run your own KB Search → Set AI → Speak cascade instead.) • Alternative: two separate projects
If the experiences really have to stay 100 % isolated, spin up two projects – each gets its own KB – and direct users to the appropriate project URL/widget based on their version. Most teams prefer the single-project + tag filter route because it keeps maintenance easier. Docs for reference
• KB Search step & filters – https://docs.voiceflow.com/docs/kb-search
• Advanced KB filtering & tagging – https://docs.voiceflow.com/docs/advanced-knowledge-base That’s all you need to safely serve version-specific answers without the AI mixing them up.
Each Voiceflow project (agent) has a single Knowledge Base. You can’t create “KB-2018” and “KB-2023” side-by-side inside the same project. • How to keep the content separate
1. Upload the 2018 files → add a KB tag like 2018
2. Upload the 2023 files → add a KB tag like 2023 • Route the user the right way
1. In the conversation ask “Which version are you using? – 2018 | 2023”.
2. Save their choice to a variable (e.g. {version}).
3. Use a KB Search step (or the KB Query API) and pass a filter such as: Only chunks that carry the selected tag will be returned, so answers never mix. (If you rely on the Agent step’s automatic KB lookup you can’t pass a filter, so just toggle the KB off and run your own KB Search → Set AI → Speak cascade instead.) • Alternative: two separate projects
If the experiences really have to stay 100 % isolated, spin up two projects – each gets its own KB – and direct users to the appropriate project URL/widget based on their version. Most teams prefer the single-project + tag filter route because it keeps maintenance easier. Docs for reference
• KB Search step & filters – https://docs.voiceflow.com/docs/kb-search
• Advanced KB filtering & tagging – https://docs.voiceflow.com/docs/advanced-knowledge-base That’s all you need to safely serve version-specific answers without the AI mixing them up.
That sounds great! I moved all entries in the KB into specific folgers, but how do I tag them? I couldnt see any option for that so far.