Explore the exciting integration of Conversational AI APIs like ChatGPT into systems with this informative presentation by Gregor Jarisch, an experienced tech entrepreneur and chatbot expert. His innovative solution, E.D.D.I - a Java-based middleware running on Quarkus, simplifies prompt and conversation management, ensuring provider independence and GDPR compliance. EDDI provides an optimized conversational AI management, enabling effective chatbot conversations and benefiting from the performance advantages of Quarkus. Learn more about the inner workings of ChatGPT bots, and discover how they can be created, deployed, and managed more efficiently with the aid of E.D.D.I.

1. Enhanced Dialog Driven Interface
Transform Workflows & Elevate Collaboration
through ChatGPT & Conversational APIs

2. Hello!
I am Gregor Jarisch
by LABS.AI
Head of R&D
Open Source Chatbot Middleware
CTO & Head of Product
Award winning Micro-Learning Solution
More about these projects:
eddi.labs.ai | gnowbe.com
2

3. 1.
What is E.D.D.I ?
Short intro
3

4. E.D.D.I – Open Source Chatbot Platform
◎ Enterprise-Certified & Cloud Native Restful Micro
Service (powered by Quarkus 3)
◎ Multiple Chatbots with Versioning for a smooth
upgrading experience
◎ Basic NLP & Behavior Rules, API Connector,
Templating
◎ Extension Oriented Plugin System
4

5. 2.
Commercial Support
Expert Knowledge Saves U-Turns
5

6. Open Source
> Core E.D.D.I
> NLP, Rules, Output,
API Connectors, …
> Full-fledged API
> Docker Image
> Optimized for Kubernetes
& Red Hat Openshift
Open Source +
Membership Subscription
Commercial
+ Dedicated Support
+ Training / Consulting
+ Customization
+ Roadmap
+ Managed Service with
24/7 Support
6

7. 25k
Lines of Code
5
Development Iterations from Scratch
15
Years of Engineering
7

8. Our Philosophy
Flexibility /
Adaptability
Structure /
Reusability
Performance /
Efficiency
8

9. 3.
Example in Action
Let’s Build a ChatGPT powered Bot
9

10. User Interaction Process
Behavior
10
Memory
Parser
Output
API
Connector

11. User Interaction Process
Behavior
11
Memory
Parser
Output
Meaning
Actions
Input
Bot Reply
API
Connector
Actions

12. User Interaction Process
Behavior
12
Memory
Parser
Output
Meaning
Actions
Input
Bot Reply
API
Connector
Actions
“<User input for ChatGPT>”

13. User Interaction Process
Behavior
13
Memory
Parser
Output
Meaning
Actions
Input
Bot Reply
API
Connector
Actions
“<User input for ChatGPT>”
trigger(send_message)

14. User Interaction Process
Behavior
14
Memory
Parser
Output
Meaning
Actions
Input
Bot Reply
API
Connector
Actions
“<User input for ChatGPT>”
trigger(send_message)
fetch_answer

15. User Interaction Process
Behavior
15
Memory
Parser
Output
Meaning
Actions
Input
Bot Reply
API
Connector
Actions
“<User input for ChatGPT>”
trigger(send_messge)
use_ChatGPT_api
fetch_answer

16. User Interaction Process
Behavior
16
Memory
Parser
Output
Meaning
Actions
Input
Bot Reply
API
Connector
Actions
“<User input for ChatGPT>”
trigger(current_message)
use_ChatGPT_api
output_answer
fetch_answer

17. User Interaction Process
Behavior
17
Memory
Parser
Output
Meaning
Actions
Input
Bot Reply
API
Connector
Actions
“<User input for ChatGPT>”
trigger(current_message)
use_ChatGPT_api
output_answer
fetch_answer
<answer from ChatGPT>

18. 4.
Another Example in Action
Let’s integrate a different API
without using ChatGPT
18

19. What’s the weather?
19

20. What’s the weather?
20
Which city would you like to know
the weather of?

21. New York
21
The current weather situation in
New York is light rain at 18°C.

22. User Interaction Process
Behavior
22
Memory
Parser
Output
API
Connector

23. User Interaction Process
Behavior
23
Memory
Parser
Output
Meaning
Actions
Input
Bot Reply
API
Connector
Actions

24. User Interaction Process
Behavior
24
Memory
Parser
Output
Meaning
Actions
Input
Bot Reply
API
Connector
Actions
“What’s the weather”

25. User Interaction Process
Behavior
25
Memory
Parser
Output
Meaning
Actions
Input
Bot Reply
API
Connector
Actions
“What’s the weather”
trigger(current_weather)

26. User Interaction Process
Behavior
26
Memory
Parser
Output
Meaning
Actions
Input
Bot Reply
API
Connector
Actions
“What’s the weather”
trigger(current_weather)
fetch_weather,
current_weather

27. User Interaction Process
Behavior
27
Memory
Parser
Output
Meaning
Actions
Input
Bot Reply
API
Connector
Actions
“What’s the weather”
trigger(current_weather)
fetch_weather,
current_weather
fetch_weather

28. User Interaction Process
Behavior
28
Memory
Parser
Output
Meaning
Actions
Input
Bot Reply
API
Connector
Actions
“What’s the weather”
trigger(current_weather)
fetch_weather,
current_weather
current_weather
fetch_weather

29. User Interaction Process
Behavior
29
Memory
Parser
Output
Meaning
Actions
Input
Bot Reply
API
Connector
Actions
“What’s the weather”
trigger(current_weather)
fetch_weather,
current_weather
current_weather
fetch_weather
“The current weather situation in
New York is light rain at 18°C.”

30. 4.
JSON to Bot
How to configure Bots in EDDI
30

31. Dictionary
31
{
"words": [
{
"word": "weather",
"exp": "trigger(current_weather)"
}
],
"phrases": [
{
"phrase": "whats the weather",
"exp": "trigger(current_weather)"
}
]
}

32. Dictionary
32
{
"words": [
{
"word": "weather",
"expressions": "trigger(current_weather)"
}
],
"phrases": [
{
"phrase": "whats the weather",
"expressions": "trigger(current_weather)"
}
]
}

33. Behavior Rule (1)
33
{
"behaviorGroups": [
{
"name": "Questions about weather",
"behaviorRules": [
{
"name": "Ask for City",
"actions": [
"ask_for_city"
],
“conditions": [
{
"type": "inputmatcher",
“configs": {
"expressions": "trigger(current_weather)"
}
}
]
}
]
}
]
}

34. Behavior Rule (1)
34
{
"behaviorGroups": [
{
"name": "Questions about weather",
"behaviorRules": [
{
"name": "Ask for City",
"actions": [
"ask_for_city"
],
“conditions": [
{
"type": "inputmatcher",
“configs": {
"expressions": "trigger(current_weather)"
}
}
]
}
]
}
]
}

35. {
"behaviorGroups": [
{
"name": "Questions about weather",
"behaviorRules": [
{
"name": "Current Weather in City",
"actions": [
“fetch_current_weather",
"current_weather_in_city"
],
“conditions": [
{
"type": "inputmatcher",
“configs": {
"expressions": "trigger(current_weather)",
"occurrence": "lastStep"
}
}
]
}
]
}
]
}
Behavior Rule (2)
35

36. {
"behaviorGroups": [
{
"name": "Questions about weather",
"behaviorRules": [
{
"name": "Current Weather in City",
"actions": [
"fetch_current_weather“,
"current_weather_in_city"
],
“conditions": [
{
"type": "inputmatcher",
“configs": {
"expressions": "trigger(current_weather)",
"occurrence": "lastStep"
}
}
]
}
]
}
]
}
Behavior Rule (2)
36

37. API Connector
37
{
"targetServer": "api.openweathermap.org/data/2.5/weather",
"httpCalls": [
{
"name": "currentWeather",
"saveResponse": true,
"responseObjectName": "currentWeather",
"actions": [
“fetch_current_weather"
],
"request": {
"method": "get",
"queryParams": {
"APPID": "<APP-TOKEN-ID>",
"units": "metric",
"q": "[[${memory.current.input}]]"
}
}
}
]
}

38. API Connector
38
{
"targetServer": "api.openweathermap.org/data/2.5/weather",
"httpCalls": [
{
"name": "currentWeather",
"saveResponse": true,
"responseObjectName": "currentWeather",
"actions": [
“fetch_current_weather"
],
"request": {
"method": "get",
"queryParams": {
"APPID": "<APP-TOKEN-ID>",
"units": "metric",
"q": "[[${memory.current.input}]]"
}
}
}
]
}
Templating Engine:

39. API Connector
39
{
"targetServer": "api.openweathermap.org/data/2.5/weather",
"httpCalls": [
{
"name": "currentWeather",
"saveResponse": true,
"responseObjectName": "currentWeather",
"actions": [
“fetch_current_weather"
],
"request": {
"method": "get",
"queryParams": {
"APPID": "<APP-TOKEN-ID>",
"units": "metric",
"q": "[[${memory.current.input}]]"
}
}
}
]
}
Templating Engine:

40. Output Templating
40
{
"outputSet": [
{
"action": "current_weather_in_city",
"outputs": [
{
"type": "text",
"valueAlternatives": [
"The current weather situation of [[${memory.current.input}]] is
[[${memory.current.httpCalls.currentWeather.weather[0].description}]] at
[[${memory.current.httpCalls.currentWeather.main.temp}]] °C"
]
}
]
}
]
}
Templating Engine:

41. Output Templating
41
{
"outputSet": [
{
"action": "current_weather_in_city",
"outputs": [
{
"type": "text",
"valueAlternatives": [
"The current weather situation of [[${memory.current.input}]] is
[[${memory.current.httpCalls.currentWeather.weather[0].description}]] at
[[${memory.current.httpCalls.currentWeather.main.temp}]] °C"
]
}
]
}
]
}
Templating Engine:

42. What’s the weather in New York?
42
The current weather situation in
Vienna is light rain at 18°C.

43. 5.
Possibilities to Code
with & around EDDI
Different Entry Points into EDDI
43

44. Restful 1: Making EDDI use of Restful API
Your Restful API
EDDI
Bot Output
User Input
Defined Request
in HttpCalls
JSON Response
from API

45. Restful 2: Send Conversation Memory to
other Service
Your Rest Service
(Receiving, Changing and Replying
the altered Conversation Memory)
EDDI
Bot Output
User Input
Callback of
Conversation
Memory
Chanced
Conversation
Memory (e.g.
filled with data
from database)

46. Java: Write a Plugin
◎ Extend NLP to extend the recognition of
user inputs
○ ai.labs.eddi.modules.nlp.extensions.dictionaries.IDictionary
○ ai.labs.eddi.modules.nlp.extensions.corrections.ICorrection
○ ai.labs.eddi.modules.nlp.extensions.normalizers.INormalizer
◎ Make new Conditions for Rules
○ ai.labs.eddi.modules.behavior.impl.conditions.IBehaviorCondition
◎ Create new LifecycleTask (like NLP, Rules, Connector, etc)
○ ai.labs.eddi.engine.lifecycle.ILifecycleTask

47. E.D.D.I
“Spin up your own instance of EDDI
and create a ChatGPT powered Bot within a few minutes”
Project Links:
Docs: docs.labs.ai
Website: eddi.labs.ai
GitHub: github.com/labsai/eddi
RH Marketplace:
marketplace.redhat.com/en-us/products/labsai