This action forms part of ThinkAutomation’s broader AI automation capabilities, enabling automated workflows to index and search large collections of text and retrieve relevant information for use in AI conversations, business processes and other workflow actions.This action enables you to store any number of text records against a Collection Name. Each record is saved with a key value. Full text searches can then be performed against the collection. Matching keys and/or text will be returned sorted by search rank. The full text database uses a full text index, allowing for fast retrieval against a large number of records.

Collection Name

Key/text pairs are contained within a Collection. Multiple collections can be used. Collection names can contain letters or numbers only. Key/text pair collections are global to the ThinkAutomation instance (IE: The same collection can be used on all Solutions/Automations).

Operations:

Set

Add a key/text pair within the Collection. Specify the Key and Text (both can contain %variable% replacements). The Assign To variable will receive the key value (or blank if an error occurred). You can use the Add Now button to manually add key/text pairs.

Search

This operation allows you to perform a full text search. Specify the search text in the Search For entry. This can be a %variable%.

The search for text can be multiple words or phrases and can contain boolean operators. Examples:

Search Text	Results
pricing quote	Searches for any text containing ‘pricing’ and ‘quote’ anywhere in the text.
“pricing quote”	Searches for any text containing the phrase ‘pricing quote’.
pricing + quote	Searches for any text containing the phrase ‘pricing quote’.
quot*	Searches for any text beginning with ‘quot’.
pricing NOT quote	Searches for any text containing ‘pricing’ and not ‘quote’.
pricing OR quote	Searches for any text containing ‘pricing’ OR ‘quote’.
pricing AND (quote OR quotation)	Searches for any text containing pricing and either quote or quotation.

From the Return list, select:

• Keys : to only return the keys
• Text : to only return the text
• Keys, Text : to return both

If text or keys+text is selected, then you can enable the Highlight Matched Text option. If this option is enabled then matched words in the returned text will be enclosed in <b> and </b> tags.

From the Return As list, select:

• Text : to return text. If ‘text’ or ‘keys, text’ is selected for the Return type, then each record will be separated by a single line ‘—‘. When only returning keys, then each key will be returned on a separate line.
• Json : to return Json data.

The Limit entry is used to specify the maximum number of records to return.

Records will be returned in rank order, with the closest matches being returned first.

The Assign To variable will receive the returned results. Use the Test button to test searches.

Delete

This operation will delete a full text record. Specify the Key to delete. The Assign To value will contain the deleted key, or blank if no record was deleted.

Drop

This operation will delete the entire collection.

Use Full Text Search with AI and RAG

Full Text Search can be used to retrieve relevant business information before passing the results to an AI model as additional context. For example, an automation could search indexed documents, emails or business records for matching information and then use the results within the Ask AI action to generate a context-aware response.

This provides another approach to Retrieval-Augmented Generation (RAG), particularly where exact words, phrases, reference numbers or structured information need to be found. Full Text Search can also be combined with ThinkAutomation’s Embedded Knowledge Store and Vector Database depending on the type and volume of information being searched.