Learn how to generate responses with your own provided data not found on your website!
Last updated
Learn how to generate responses with your own provided data not found on your website!
Last updated
If you'd like to use generated answers, but critical data is missing from your website and/or in a format that cannot be read by our web-scraping tool (PDFs, images, videos), then you will want to create a documented response!
Documented data is pulled from uploaded documents in the dashboard and it reads the same way as scraped URL data. This data can be adjusted directly within the dashboard.
To add new documented data, follow the steps below:
While in the Satisfi Dashboard, go to Studio -> NLP Manager -> Responses
Under LLM Content Group, find and Open the "doc_sample_response" name
Click on the Clone button
Create a new response name using doc_[topic]
Follow the guidelines specified for adding content, such as:
Add the topic header
Add relevant content
Use a 5-dash separator between topics if multiple topics are within one response
Hit Publish
Training will automatically occur to update your chat
You can have any number of cloned doc_responses divided by categories of content, like policies, tickets, kids, food, and much more
To update documented data, follow the steps below:
While in the Satisfi Dashboard, go to Studio -> NLP Manager -> Responses
Locate the corresponding response labeled as “doc_” and click on it
Hit Edit and update the response
When done, press Ready to Publish -> Publish
To delete documented data, follow the steps below:
While in the Satisfi Dashboard, go to Studio -> NLP Manager -> Responses
Locate the corresponding response labeled as “doc_” and click on it
Click the trash can icon and confirm
Since documented data is used to create generated responses, you are unable to add enrichments directly in documented data. To learn how to add enrichments to generated responses, click the link below!
We provide a sample of how your documented data should be formatted in your LLM volume. You can find the example in your library under the response name "doc_sample_response". Below are a few formats we recommend when adding documented data:
-----
The following inputs are related to buying tickets: How do I buy tickets? Can I purchase tickets? Are tickets available? Answer: Yes, tickets are available for purchase on the website.
-----
Season Tickets: When do single-game tickets go on sale? Answer: Single-game tickets go on sale at the beginning of the season.
-----
Parking Information:
Parking lots open four hours before the first pitch Parking is first come, first serve and is subject to availability. Parking is $15 in all lots.
When cloning the sample documented response name (doc_sample_response), be sure to remove any sample information listed and review your data before publishing!
We always recommend that you review responses that are generated and exposed within your chat.
If your documented data is not being understood by the LLM, this may be due to poor formatting within the response. To fix this:
Find the corresponding labeled response in the library
Click Edit
Ensure that:
There is a header description related to the data's topic
All sections are separated by a 5 “-----”
No section is very short or extremely long