conscious-sapphire
conscious-sapphire15mo ago

Unable to find relevant answer

Anyone knows why this is happening? The same question even though it knows the answer some time it responds as unable to find relevant answer. Then if asked again it will provide the answer. Any solutions?
No description
10 Replies
Tico | Voiceflow Assistant
To implement the Knowledge Base in Voiceflow, you can follow these steps: 1. Adding Data Sources: Navigate to the Knowledge Base tab within the Content section of your dashboard. Click the 'Add Data Source' button and select the type of data source that aligns with your requirements. Options include URL(s), Plain text, Sitemap, PDF, and Docx. 2. Setting Refresh Rate: For each data source, you can set a refresh rate ('Never', 'Daily', 'Weekly', or 'Monthly') to ensure the information remains up-to-date. This is particularly useful for ensuring that your AI responses are always based on the latest available data. 3. Customizing Settings: In the KB Settings modal, you can adjust settings such as Model, Temperature, Max Tokens, Instructions, System, and Chunk Limit to tailor the generation of responses to your specific needs. 4. Previewing Responses: Use the 'Preview' feature to test your Knowledge Base. Enter a question in the 'Question' field and click 'Send'. Review the response and the sources listed under the 'Sources' section to verify that the AI is pulling data from the correct segments. 5. Managing Integrations: If you are using integrations like Zendesk, ensure that the connections are active and correctly configured. Any disruption in the connection can affect the refresh rate and data retrieval. For more detailed guidance, you can refer to the following resources: - How the Knowledge Base Works - Step-by-Step Guide to Creating an Assistant
conscious-sapphire
conscious-sapphireOP15mo ago
Also for the same question, it might give 2 completely different answers. Even though I have temperature at 0.1
Steve B. | IntelleSync.ai
I presume your using knowledge base? What documents are using in your knowledge base, how are you providing the information? What AI model are you using?
conscious-sapphire
conscious-sapphireOP15mo ago
I've input them as plain text, and I'm using gpt 3.5 turbo so I only have plain text currently in my knowledge base. I've noticed that this is something that happens when I ask a lot of questions, it's like the bot time outs? is this something that might happen?
Steve B. | IntelleSync.ai
Ok cool, how are you formatting the question for your knowledge base?
conscious-sapphire
conscious-sapphireOP15mo ago
Each plain text is a section then titles, subtitles and content
Steve B. | IntelleSync.ai
OK, the best way to have plain text is Q: & A: and then converted to markdown format, so it would look something like this: Q: What AI services do you provide? what services do you provide? A: We provide chatbots and voice agents. then converted to markdown:
**Q:** What AI services do you provide? What services do you provide?

**A:** We provide chatbots and voice agents.
**Q:** What AI services do you provide? What services do you provide?

**A:** We provide chatbots and voice agents.
I would do this for each set of questions and answers. The more you group info the harder it is to chunk. Apparently LLMs love markdown.
conscious-sapphire
conscious-sapphireOP15mo ago
I understand, but I’m trying to build a legal assistant, and Q&A’s are not structured like this.
Steve B. | IntelleSync.ai
You can add titles, subtitles and bullet points in your answers if that’s the issue
conscious-sapphire
conscious-sapphireOP15mo ago
Yes that’s how it is currently

Did you find this page helpful?