Released: Jun 9, 2025

Mises à jour de v5.0.1065

Fonctionnalités

• Added "Lookup From A Database Using AI" action. This action enables you to query a database using natural language. The database schema along with the natural language query is sent to AI to generate a parameterized SQL select statement that is then executed and the results returned to a variable.
• Added "Embedded Vector Database" action. ThinkAutomation now includes a built-in vector database for storing large numbers of items for use with the "Ask AI" action or for performing custom semantic search operations.
• Added "Secret Constants" to the "Global Constants". This allows global %variables% to be assigned where the value is securely stored in the Windows Credential Manager. Secret values are not saved with Automations and will not appear in any logs.
• Added GPT-4.1 and GPT-4.1-mini models to the OpenAI AI Provider type. These models have a much larger maximum tokens size (1,047,576 tokens) allowing a larger context window.
• Added support for PDF files in addition to images for the "Ask AI" action with the "Ask AI To Respond To A Prompt With An Image" operation.
• Added "Include Sub-Folders" to the "File Operation - Get Folder Contents" operation. The list of files can now be returned as text (comma or line separated) or a JSON array.
• Added support for sub-folders and multiple file masks (e.g. *.pdf, *.docx) to the "File Pickup" message source type.
• Improved the "Database" message source type: it will now reset the connection pool on failed connections. This will improve reconnections when the database comes back online.
• Improved the "File Pickup" new message source wizard so it will now ask if you want to add document text content to a Vector Database for use with AI. It will also ask if you want to create a chat message source and automation. This enables you to quickly create an AI based conversational front-end to a local file-based document store.
• Added "PDF Use OCR For Image Only Documents" option to the "Convert Document To Text" action. This is useful when converting multiple PDF documents to plain text and some of the PDF documents contain only scanned images.
• Added "Auto Rotate" and "Page Segmentation Mode" options to the "Convert Image To Text Using OCR" action.
• Added new method for HTML to plain text conversion. The new method preserves more layout and will convert HTML data tables to aligned rows and columns in the plain text. The new method is not enabled by default on existing ThinkAutomation installs. It must be enabled in Server Settings - Message Processor. The differences can be tested using the "Text Operation" action preview.
• Added "Extract - Summarized Text" operation to "Text Operation" action. Summarizes text into the most relevant sentences.
• Added "Tokenize Into Sentences" option to the "Tokenize" action. It returns the top x sentences for any text based on keyword density.
• Added option to "Email" message source types to restrict processing to messages received in the last x days.
• The "Update A Database Using CSV Or JSON" action now allows you to create a new table. This is useful if you need to quickly a new table prior to importing CSV or JSON.
• Improved PDF to plain text conversion results when PDF documents contain only images.