Ai connect pro_03_Email Analysis_ChatGPT
Flow
Created at 1/28/2025 8:32 AM
Ai connect pro_03_Email Analysis_ChatGPT
232
No
43
E:/BlueWest/98-Utilitaires/Grille Switch.png
E:/BlueWest/bitBucketFolder/ai-connect-pro/Images/Ai-connect-+-gpt.png
C:/Users/tdesc/Pictures/gpt.png
No
No
00:00
00:00
No
Monday
Sunday
No
1
Start of the month
282
Folder
folder
-
#automanaged#
No
Default
No
00:00
00:00
No
Monday
Sunday
No
1
Start of the month
#808080
No
No
Yes
Yes
Overwrite
1
No
No Folders
No
560
304
ProblemFilesFolder
producer
Filter
One
Problem jobs
This flow only
Default
10
10
80
224
Folder
folder
-
#automanaged#
No
Default
No
00:00
00:00
No
Monday
Sunday
No
1
Start of the month
#808080
No
No
Yes
Yes
Overwrite
1
No
No Folders
No
176
304
SubmitPoint
producer
Filter
Unlimited
EML
#automanaged#
Default
<p>This flow shows how Enfocus Switch can use AI to automatically process incoming emails related to print production.</p>
<ul>
<li><strong>Email content and attachments</strong> are analyzed using a custom prompt</li>
<li>Information is extracted and structured into a <strong>clean XML format</strong></li>
</ul>
<div class="highlight">
The XML is used to classify the email and route it to the correct department or workflow automatically.
</div>
<ul>
<li>Job and order details (file name, OS, delivery errors, software used)</li>
<li>Email metadata (sender, subject, date, priority, attachments)</li>
<li>Category of request (e.g., invoice, quotation, delivery issue)</li>
<li>Suggested summary and response for customer service</li>
</ul>
<h2>How It Works</h2>
<p>
When an email arrives (e.g., to <strong>contact@myprintshop.com</strong>), the system parses it and feeds it to an AI prompt.
The output is a structured XML file that Switch uses to route the request efficiently.
</p>
<h2>Prompt Logic</h2>
<p>
The prompt is crafted to:
</p>
<ul>
<li>Classify the type of request from the email content</li>
<li>Extract structured metadata and order info</li>
<li>Infer missing values when possible</li>
<li>Generate a helpful summary and a draft reply</li>
</ul>
<p><strong>This flow eliminates manual sorting and accelerates job handling in customer-facing print operations.</strong></p>
E:/BlueWest/bitBucketFolder/ai-connect-pro/Images/AI Connect_200px.png
No
No
No
No
Job only
Files only
<ValueDescription Type="stringlist">
<Value>eml</Value>
</ValueDescription>
Submit
Yes
No
80
304
SwitchClient
Filter
90
119
93
Gray
No
All Files
No Files
Folder
folder
GPT response
#automanaged#
No
Default
No
00:00
00:00
No
Monday
Sunday
No
1
Start of the month
#808080
No
No
Yes
Yes
Overwrite
1
No
No Folders
No
560
224
Folder
folder
-
#automanaged#
No
Default
No
00:00
00:00
No
Monday
Sunday
No
1
Start of the month
#808080
No
No
Yes
Yes
Overwrite
1
No
No Folders
No
368
304
Filter
0
2
179
Gray
No
All Files
No Files
createLog
processor
TrafficLight
Create log
656
304
Default
Default
Yes
Unlimited
XML
[Metadata.Text:Dataset="AIconnect",Model="JSON",Path="/0/message/content",Space="trim"]
####
%%%%
utf8
Folder
folder
Data
#automanaged#
No
Default
No
00:00
00:00
No
Monday
Sunday
No
1
Start of the month
#808080
No
No
Yes
Yes
Overwrite
1
No
No Folders
No
752
304
TrafficLight
0
179
180
Gray
Data with log
Log
No
Yes
Yes
Yes
Filter
0
93
218
Gray
No
All Files
No Files
Folder
folder
XML
#automanaged#
No
Default
No
00:00
00:00
No
Monday
Sunday
No
1
Start of the month
#808080
No
No
Yes
Yes
Overwrite
1
No
No Folders
No
752
224
TrafficLight
-90
179
212
Gray
Log
Log
No
Yes
Yes
Yes
EMLPickup
processor
Move
EMLPickup
272
304
Default
Default
Yes
Unlimited
EML
Metadata is asset
utf8
Yes
Attachments
Move
0
218
163
Gray
No
Folder
folder
Set your API key
#automanaged#
No
Default
No
00:00
00:00
No
Monday
Sunday
No
1
Start of the month
Magenta
No
No
Yes
Yes
Overwrite
1
No
No Folders
No
464
384
ai-connect-pro
processor
TrafficLight
AI connect_PRO
464
304
Default
Default
Yes
Unlimited
OpenAI
Text completion
dall-e-2
Vivid
Standard
256x256
1024x1024
a photo of a happy corgi puppy sitting and facing forward, studio light, longshot
1
gpt-4o
You are an automated agent for a print production company.
Analyze the email content and metadata below and generate a single-line XML
Instructions:
Classify the email into one category:
- Invoice
- Bill
- Quotation demand
- Quality claim
- File for order
- Delay question
- Delivery question
If none fit, use <Category name='Other'>*Proposed*</Category>
Extract or infer:
- Sender, recipient, subject
- Priority: Low | Normal | High (default: Normal)
- Attachments: list all, even if empty; add purpose as metadata
If order info is present:
- Order number/reference
- Job type/file name
- Quantity, material, size, delivery date, instructions
- Billing details if applicable
Output must:
- Output strict XML only, no plain text or line breaks, or markdown wrapper
- Match this exact structure (use single quotes, escape XML special characters, no extra fields):
<XML><Metadata><Priority>*</Priority><Attachments><Attachment><FileName>*</FileName><FileType>*</FileType></Attachment></Attachments></Metadata><Order>*Order details, one XML tag by info*</Order><Category name='*'></Category><Summary>*Summary*</Summary><Response>*Suggested reply*</Response></XML>
0.5
1.0
alloy
mp3
Current
en
0
json
http://localhost:7869/
Text completion
Tell
gemma:2b
0
Text completion
You are an expert assistant from MyPrintingCompany.com a printing company, helping with customer communication. You will receive two input files:
- An XML file that contains all information about a printed product (e.g., paper type, size, deadlines, invoice/bill info, contact details).
- A Pitstop preflight report file that details the technical results of the PDF preflight check.
Your task is to:
- Read and understand the relevant data from both files.
- Generate an HTML email body only (no subject, no headers).
Instruction for the HTML content :
- Translate it in [Metadata.Text:Dataset="Submit",Model="XML",Path="/field-list/field[tag='Language']/value"]
- Add at the top a greeting to the customer (refer to the XML file for contact)
- Follow with a reminder of the order name and a summary block (in a table format) showing key production details from the XML.
- Summarize the preflight report using clear, non-technical language intended for someone not used to printer related terminlogy.
- Errors are critical and mean the file is not suitable for print — highlight these in red.
- Warnings are advisory and up to the customer to accept — highlight these in orange.
- If there are no errors or warnings, briefly confirm the file is ready for print.
Display the preflight results in a table format, listing:
- Type (Error or Warning)
- Description (
- Page number (if available)
- Include helpful suggestions for resolving issue
- Add a button with the review link that should lead the customer to this web page : https://www.enfocus.com/en/review
- Below the button, add a short message telling the customer to click the button to make an approval decision.
Formatting :
- Output the body of the HTML only, as a single line (no line breaks, no \n or \r).
- Do not escape characters.
- Use single quotes (') instead of double quotes (").
- Do not include any plain text explanation or mardown element — only the final body HTML output.
mistral-large-2411
0
store_name=The name of the store or merchant transaction_date=Date and time of the purchase total_amount:number=Total amount paid including tax tax_amount?:number=Tax amount applied (if shown) payment_method=Method of payment (e.g., Visa, cash) address?:string=Store address if available
Yes
No
EML
No
json
No
TrafficLight
0
174
2
Gray
Data with log
AIconnect
No
Yes
Yes
Yes
TrafficLight
-90
174
148
Gray
Log
Log
No
Yes
Yes
Yes
Filter
0
156
174
Gray
No
All Files
No Files
Filter
0
163
174
Gray
No
All Files
No Files