How to create a Collection of AI Chat Packages
Home AiWerkz Home Chat Packages Home Prompt Manager Chat Package Demo Site Login Logout Sign up Create a Chat Package Upload a file to create a Chat Package Create a Lite Chat Package My Chat Packages Create a Chat Package Group My Chat Package Groups Domain Management Learning Hub Pricing About AiWerkz Privacy Terms of Use

Creating a Chat Package Collection

Collection versus Group
 
We are currently evaluating whether to discontinue collections in favor of chat package groups.
 
Scenario: Chatting with a 500 page PDF
We have a nearly 500 page PDF manual for an avionics system that we want to enable for AI chat. The document is too large to be converted into a single chat package so we created an ICE (Interactive Chat Enabled) Package Collection. There are several steps involved in creating the Collection. We decided to break the document up based on its sections. For each section we then manually created prompt packages of the same collection name.

Prepping the content for uploading into chat packages involved some manual work. After exploring various methods to automatically prepare documents for conversion to chat packages we decided to stick with a rather manual method because it provided the best results. While the following describes how to break up a large document into a chat package collection the same process applies to having several documents.

Step 1: Extract Text and Break the Document into Sections
A chat package is limited to a maximum length of 120,000 characters which is typically in the range of 20,000 to 26,000 words. If you are within that range then just create a single chat package. For our pdf with almost 500 pages it has many more words than what can fit into a single chat package. So, we break it up into pieces to become individual chat packages. We like to use the following Python script. Feel free to use it. Or break up the document entirely manually. The end result is the same; you have a bunch of text to paste into the "Create Chat Package" web page.

Python PDF Parser Script

Step 2: Construct Document Summary and Keywords
You now have several text files. Each text file represents a section or chapter of the large PDF. For each of these sections a Document Summary and Document Keywords will need to be created. Of course, you can create these manually. However, we hacve that it is easy to paste the the document section into ChatGPT with these instructions:


For the following content please create a summary of less than 2000 characters.  
Then generate a comma separated list of keywords, product names, product model 
numbers and key concepts of less than 2000 characters.
```
Content here...

Step 3: Create the Chat Packages
There will be one chat package per document section. Detailed instructions are here:
How to Create a Chat Package

The link to create a chat package is at:
Create a Chat Package

It is important to use the same "Collection ID" and "Collection Name" when creating several chat packages that are intended to be used as a collection.

Once you have created some or all of the chat packages then you will see them in My Chat Packages

In My Chat Packages you will see a link to "ICE Pack Collection". Click on it and the Collection will open. You can share that link with whomever you like.