View source for 2024:AI Analyst (Object)

{{AutoVersion}}

<blockquote>{{#lst:Glossary|AI Analyst}}</blockquote>

{|class="important-box"
|
'''!!'''
|
'''''WARNING!! AI ANALYSTS ARE AI ASSISTANTS IN GROOPER 2025 (AND BEYOND)'''''

'''AI Analysts''' are an early attempt at what would become '''[[AI Assistant]]s''' in Grooper. You will not find AI Analysts in any version of Grooper other than 2024
|}

{|class="download-box"
|
[[File:Asset 22@4x.png]]
|
You may download the ZIP(s) below and upload it into your own Grooper environment (version 2024). The first contains one or more '''Batches''' of sample documents.  The second contains one or more '''Projects''' with resources used in examples throughout this article. 
* [[Media: 2024_Wiki_AI-Analyst_Batch.zip]]
* [[Media: 2024_Wiki_AI-Analyst_Project.zip]]
|}

== About ==
Given the general open nature of how "chatting" with a document can be, specifying what exactly to do with the AI Analyst is difficult. In fact, finding creative ways to interact with your documents via a chatbot is the whole point. One great usecase for the AI Analyst would be asking for a summary of a documents if you are generally unfamiliar with them otherwise. This could be helpful for a designer wanting to know what to collect from the document, or a reviewer who needs to understand what they need to verify.
<div style="padding-left: 1.5em;">
=== Chat with Your Documents ===
[[image: 2024_AI-Analyst_00_01.png|right]]
* '''AI Analyst''' is a new node type that spins up a document chatbot.
* It defines a task-specific LLM assistant used to chat with documents.
* Selecting one or more documents, users can chat with the assistant in '''Grooper''' about the selected documents.
** Text data from the selected documents are fed into an assistant's knowledge base.
** Knowledge files are divided into chunks.
** Embeddings are stored for each chunk.
** As users enter questions, an embeddings-based search finds chunks of document content relevant to the question and feeds them into the context.
* Each chat is stored as a child Chat Session organized by user. 
<br clear=all>
</div>

== How To ==
<div style="padding-left: 1.5em;">
=== Establish an LLM Connector ===
First, we need to establish an '''''LLM Connector''''' within the '''''Options''''' property on the '''Root''' object.

Please visit the '''''[[LLM Connector (Repository Option)|LLM Connector]]''''' article for more information.

=== Configure the AI Analyst ===
# Right-click on the provided '''Project'''.
# Choose "Add > AI Analyst" from the pop-out menu.
# Name the object in the "Add" window.
# Click the "Execute" button.

[[image: 2024_AI-Analyst_01_01.png]]


# <li value=5>Click the ellipsis button for the '''''Instructions''''' property.<li>
# Add the provided text in the "Instructions" window.<pre>You are an invoice analyst.  Your job is to answer questions about the selected invoice or invoices documents. Your knowledge files contain the text content of one or more invoices. Each page of content is separated by a form feed character.  When responding, please reference source material and include short verbatim quotes which can easily be linked back to the original knowledge content.</pre>
# Click the "OK" button.

[[image: 2024_AI-Analyst_01_02.png]]


# <li value=8>Click the drop-down button on the '''''Knowledge''''' property.<li>
# Select ''Text Knowledge'' from the drop-down menu.

[[image: 2024_AI-Analyst_01_03.png]]


# <li value=10>Click the ellipsis button for the '''''Predefined Messages''''' property.</li>
# In the "Predefined Messages" window click the "Add" button. This will add a "Content" entry to the collection.
# Click the ellipsis button for the '''''Content''''' property.
# Add the provided text in the "Content" window.<pre>Summarize this document.</pre>

[[image: 2024_AI-Analyst_01_04.png]]

=== Use the AI Analyst to Chat with Your Documents ===
# Select the provided '''Batch''' from the "Batch > Test" folder.
# Click on the "Viewer" tab.
# Right-click the document from the provided '''Batch'''.
# Select "Discuss" from the drop-down menu.
# Click the drop-down button for the '''''AI Analyst''''' property.
# Select the recently created '''AI Analyst''' object.
# Click the "Execute" button.

[[image: 2024_AI-Analyst_02_01.png]]


# <li value=8>Selected documents will be displayed in the top-right of the UI. ALl text from all documents will be submitted to the AI. A selection here only affects what is shown in the "Document Viewer".</li>
# A custom prompt can be written in the bottom portion of the UI.
# Above the prompt entry are interaction controls for the chat session.
# Click the "Predefined Messages" button.

[[image: 2024_AI-Analyst_02_02.png]]


# <li value=12>Select the predefined message from the "Send Predefined Message" window.</li>
# Click the "OK" button.

[[image: 2024_AI-Analyst_02_03.png]]


# <li value=14>The selected predefined message will be inserted in the prompt area.</li>
# Click the "Submit" button to engage the AI chatbot.
# Upon submission the "User" message will be sent and displayed in the top-left portion of the UI.
# The AI response will be displayed beneath the "User" prompt.

[[image: 2024_AI-Analyst_02_04.png]]


# <li value=19>A '''Chat Session''' object will be created as a child to the '''AI Analyst''' object. These objects are user specific and only visible to the user that started the chat session.</li>
# With the '''Chat Session''' object selected you can click on the "Chat" tab to continue chatting with the AI.

[[image: 2024_AI-Analyst_02_05.png]]


# <li value=21>With a chat session started the "Chat" page can now be used.</li>
# You can select the chat session on the left side of the UI.
# You can choose to delete the chat session.
# You can also view properties of the chat session by clicking the "Properties" button.
# In the "Chat Session" window you can click the ellipsis button for the '''''Documents''''' property to open the "Documents" window.
# In the "Documents" window you can select the documents viewed and submitted in the chat session.

[[image: 2024_AI-Analyst_02_06.png]]
</div>