Calculate Embedding Action:
An embedding is numerical representation of a text or data. The text provided is converted into floating point numerical form.
This is used during search. So search string is assessed and it is determined how close it is to the embeddings.
A short distance between embedding and search string implies that meanings match.
Embeddings can be generated and used as follows
Calculate Embeddings Action in Designer workflow:
Calculate Embeddings Action in Legacy workflow:
General Settings:
Action Name: enter a name for Calculate Embeddings action
Action Description: enter description for Calculate Embeddings action
| Delay Execution: it will delay action execution for specified seconds. |
Modify item(s) with system update: select this option if you want to modify item(s) with system update.
Embeddings Settings:
AI Connection: Click arrow to navigate to 'Crow Canyon NITRO Settings' page to create 'AI Connection Settings' if not exist already.
Click 'New AI Connection' button to create a new connection
Contact your organization for AI Connection Settings details.
After filling details, test connection to verify connectivity with AI
Save the settings
Search Text Column(s): select column(s) in which you want to search data.
Embeddings Column: select Multi-line text column to store embeddings value. Embedding model is used to generate these Embedding values. This model is defined in 'AI Connection Settings'.
MD5 Checksum Column: select Single-line text column to store Checksum of search text column(s). Checksum is used to detect changes in the content.
even minor change in content will generate completely different checksum.
Example of This calculate Embedding:
In Helpdesk Application, Embedding is calculated for Knowledge base articles.
In KB co-pilot, when search string is passed, based on the nearest match in Knowlege base articles, result will be provided.
Workflow: Below workflow is configured on Knowledge Base list item(s) update event
AI Connection Settings:
Ask organizaton for API Key
when knowledge base article is updated, Embeddings and Checksum is calculated as shown below:
KB Co-Pilot : In Helpdesk Application, click 'KB Co-Pilot' in top link bar. Enter search string in box, it will give as output the nearest matching result from Knowlege Base articles using Azure Open AI.
