Download "Why abstractions are useful (storing openai conversations into a db)"

Please wait. We're preparing links for easy ad-free video watching and downloading.

Previous video: Android Studio SQLite Tutorial Retrieve data Display in a list view Next video: How to use NEW Input System Package! (Unity Tutorial - Keyboard, Mouse, Touch, Gamepad)

How to use NEW Input System Package! (Unity Tutorial - Keyboard, Mouse, Touch, Gamepad)

How to use NEW Input System Package! (Unity Tutorial - Keyboard, Mouse, Touch, Gamepad)

Channel: Code Monkey

Android Studio SQLite Tutorial Retrieve data Display in a list view

Android Studio SQLite Tutorial Retrieve data Display in a list view

Channel: Channel Ai

web development

programming

coding

code

learn to code

tutorial

software engineering

00:00:01

around with this text Adventure RPG that

00:00:03

uses open Ai and also something called

00:00:05

Lang chain to basically keep track of

00:00:07

the conversation I'm having with the AI

00:00:09

system so that when I ask it questions

00:00:12

in the future it can kind of go back and

00:00:13

check what has already been discussed

00:00:15

what is the actual scenario for this

00:00:18

adventure that we're going on and when I

00:00:20

first started building this I was using

00:00:21

something called buffer memory to store

00:00:23

the information inside of the memory of

00:00:26

the computer right the issue with that

00:00:27

is that when I stop the program it

00:00:30

clears Out All the known memory right so

00:00:32

you don't know what your character is

00:00:34

you don't know where you are in the

00:00:35

actual dungeon you don't know how many

00:00:36

hit points you have it all kind of just

00:00:38

gets wiped away right and that makes

00:00:39

sense so I've been looking into a way to

00:00:40

basically store the history or the

00:00:42

memory inside an external database so

00:00:44

that when I kick this off again it's

00:00:46

going to continue off where we left off

00:00:48

so with that background information all

00:00:49

being discussed the reason I'm making

00:00:52

this video is I want to discuss

00:00:53

something that's very important when it

00:00:54

comes to programming and that is

00:00:55

interfaces you are using interfaces all

00:00:58

the time we declare functions you need

00:00:59

to clear arguments in the return values

00:01:01

of those functions and you may not know

00:01:04

about what the interface is if you're a

00:01:05

beginner interfaces are a great way to

00:01:08

allow you to extend and swap out

00:01:10

functionality just by using like

00:01:12

abstract classes and other types of

00:01:14

approaches right so let me show you what

00:01:15

I did so what do I mean by that right

00:01:17

previously we're using buffer memory and

00:01:19

I'm taking this variable and passing it

00:01:20

to this memory argument of conversation

00:01:23

chain right if you hover over this

00:01:25

you'll notice that memory has to be a

00:01:28

base memory type of class so what this

00:01:31

allows us to do is that whenever you

00:01:33

declare a new class as long as it

00:01:35

extends

00:01:36

some existing functionality like an

00:01:39

abstract class or something like that

00:01:40

and if you keep diving down the

00:01:43

extension chain so like you'll notice

00:01:44

that this one extends base memory what

00:01:46

this allows you to do is you can start

00:01:48

declaring your own classes and because

00:01:50

they match the expecting interface of

00:01:53

this memory property everything will

00:01:55

just work right so in order to extend

00:01:58

instead of using the in memory buffer

00:02:00

memory I can use my own external memory

00:02:02

all I need to do is basically make my

00:02:05

own generic chat memory and notice that

00:02:08

this is just a class that extends base

00:02:09

chat memory and when you extend an

00:02:12

abstract class most of your Ides are

00:02:14

going to give you like suggestions of to

00:02:16

like as to like what you need to

00:02:17

implement for example if I comment all

00:02:18

this out you'll notice that this is red

00:02:20

saying that you were not implementing

00:02:22

the correct abstract members so if I go

00:02:24

down here and say Quick Fix notice that

00:02:27

it'll add all the methods that I need to

00:02:29

basically get started to enable this to

00:02:32

store memory in an external database

00:02:34

right so I added those required methods

00:02:37

and basically when the black box which

00:02:40

is this conversation chain decides that

00:02:42

it needs to either load or write the

00:02:45

conversation history it's going to call

00:02:47

one of my two methods here it's either

00:02:48

going to call load memory variables or

00:02:51

it's going to call Save context and

00:02:53

because I follow the interface it's very

00:02:55

easy to kind of like extend the

00:02:57

functionality just by taking these

00:02:58

inputs writing them to a database or

00:03:00

taking these inputs and fetching them

00:03:02

from a database again this video is more

00:03:04

for like beginners so if you're kind of

00:03:05

advanced you already know about Audrey

00:03:06

Dorney programming and Abstract classes

00:03:08

and interfaces this is going to be like

00:03:09

a boring video but overall the point I'm

00:03:12

trying to make is that like all I needed

00:03:13

to do was make a new class that extends

00:03:15

that abstract class and then I pass it

00:03:18

my own methods for writing and loading

00:03:20

history from my own type of data stores

00:03:22

so in this case I am storing and

00:03:24

retrieving the history data in a

00:03:25

third-party service called convex which

00:03:28

is basically a backend as a service and

00:03:29

it also gives you the database without

00:03:31

having to set that up as well and notice

00:03:33

that when at some point conversation

00:03:35

chain decides to load the history all

00:03:39

I'm saying is query my database call

00:03:41

this load history method and then join

00:03:44

all of the conversation histories into a

00:03:47

single string and return that right so

00:03:49

all this thing needs to do is just

00:03:50

return a string of the previous chat

00:03:52

conversations together and this thing

00:03:55

can figure out how to like do what it

00:03:57

needs to do all right same thing with

00:03:58

writing history events I get an input

00:04:00

and a response and I just need to write

00:04:02

that somewhere to doesn't matter what

00:04:03

database I write that to this could be a

00:04:05

file store this could be somewhere else

00:04:06

but this thing was written in a good way

00:04:08

where like it depends on a nice

00:04:10

interface that allows you to kind of

00:04:11

swap that out whenever you need to so

00:04:13

let me just run this real quick I'm

00:04:15

gonna say mpm Run start

00:04:17

okay and so that's gonna basically you

00:04:19

know set me up with like you are a

00:04:20

warrior you're going to go into a

00:04:23

haunted place and then it's asking are

00:04:25

we ready to begin I'm going to say yes

00:04:27

we are ready like this all right a group

00:04:29

of Undead creatures approaches you with

00:04:31

their eyes glowing with the Eerie light

00:04:32

they raise their swords really 20-sided

00:04:34

die to determine your attack roll let's

00:04:35

just go and roll one real quick I'll do

00:04:37

a nine okay and it kind of tells you

00:04:39

that you hurt one of the undead

00:04:41

creatures and then they also strike me

00:04:42

the point I'm going to show here is that

00:04:44

uh if you go up to your database now

00:04:46

notice that that history has been stored

00:04:49

inside a table here where I have the

00:04:51

chronological order of all the inputs

00:04:54

and the responses from open Ai and

00:04:57

behind the scenes when lingchain decides

00:04:59

that it needs to like tack on another

00:05:01

call it's just using this database under

00:05:04

the hood right all I had to do is extend

00:05:05

the interface and I'm like good to go so

00:05:07

I guess what I'm trying to say is that

00:05:09

it is important to understand how

00:05:10

implicit and explicit interfaces work in

00:05:13

software I mean you use them all the

00:05:15

time whether you know about them or not

00:05:16

and using something like typescript or

00:05:18

Java or strongly typed static languages

00:05:20

allows you to have an explicit interface

00:05:23

to find so that people can come by and

00:05:25

easily extend your lower level Library

00:05:27

if they need to right so the better you

00:05:29

define your interface the better your

00:05:32

abstraction could potentially be alright

00:05:34

so now that I kind of got that out of

00:05:35

the way I do want to talk about some of

00:05:36

the hurdles I'm running into with this

00:05:38

project one thing is every time that you

00:05:40

basically you know roll a dice or send

00:05:42

in a command this has to take all of the

00:05:45

previous conversation history and it

00:05:47

sends it the open AI so this

00:05:48

continuously gets larger and larger and

00:05:50

larger as the game goes on and at some

00:05:52

point after like five or six responses

00:05:55

that you type in the token length gets

00:05:57

about a thousand now I think open AI

00:05:59

with this turbo model charges like 0.00

00:06:02

one five

00:06:04

Cent

00:06:05

per 1000 tokens so basically this is

00:06:08

going to start adding up right you're

00:06:09

going to keep on getting charged as your

00:06:10

gain gets longer and longer and also

00:06:12

you're going to hit limitations there's

00:06:13

only so many input tokens that you can

00:06:15

provide these models some some are set

00:06:17

to 4 000 some are 8 000 some have a much

00:06:20

higher threshold but they have different

00:06:22

costs associated with them right so one

00:06:24

thing I started adding into this game

00:06:25

was basically after you have X amount of

00:06:28

prompts that get run what we're going to

00:06:30

do is we basically summarize the state

00:06:32

of the game and we have openai basically

00:06:35

print out that summary somewhere let me

00:06:36

just go ahead and roll um again let me

00:06:38

roll 20. at some point it should

00:06:40

summarize and it should print out how

00:06:42

much help I have what's in my inventory

00:06:45

um here it is okay so all right so what

00:06:47

I'm doing after like five or six prompts

00:06:49

is I'm telling open AI to please

00:06:50

summarize the game State including the

00:06:53

goals the current location player Health

00:06:55

items into a single paragraph so we can

00:06:57

easily continue at a later time please

00:06:58

use the least amount of words as

00:06:59

possible okay so what this does is it

00:07:01

basically prints out the state it says

00:07:04

the player is currently engaged in a

00:07:06

battle with Undead creatures they have

00:07:07

successfully landed hits on two of the

00:07:09

creatures dealing six points of damage

00:07:10

each the player has also managed to

00:07:11

block two attacks take no damage there

00:07:14

are still two Undead creatures remaining

00:07:15

the player's current location is unknown

00:07:17

and their health and inventory are not

00:07:19

specified again why I'm doing all this

00:07:20

is so that if someone were to decide to

00:07:22

close the game and come back it's going

00:07:24

to load in the chronological history

00:07:25

it's going to hopefully read in one of

00:07:27

these summarize game States and open AI

00:07:30

is going to be able to basically resume

00:07:32

where you left off right so it's only

00:07:34

going to grab the past like five prompts

00:07:36

so these things will be emitted so if

00:07:38

there's anything that's super critical

00:07:39

that was omitted here and that's not

00:07:41

included in your summary then you're

00:07:42

going to actually like have a weird

00:07:43

experience they kind of have to play

00:07:45

around with like how much memory you

00:07:47

want to give this thing so in this case

00:07:49

I give it five but maybe you want to

00:07:50

give it 10 or something again the higher

00:07:53

the memory the more this will cost to

00:07:55

keep playing what I do is in the game

00:07:57

first initializes I check

00:07:59

is this the first time we've played if

00:08:01

not then we ask openai remind me where

00:08:04

we left off okay and if you notice here

00:08:06

it's going to use that history that's

00:08:07

stored in the database and it's going to

00:08:09

say you're currently engaged in battle

00:08:10

with Undead creatures it basically just

00:08:12

printed out the very last summary but um

00:08:15

this could potentially there could be

00:08:17

more stuff above here that like needs to

00:08:18

be summarized as well so I'm going to go

00:08:20

ahead and say attack the

00:08:23

creatures so now there's goblins

00:08:25

attacking me those games are getting

00:08:26

more intense and what I did also do is

00:08:28

down here I have a math.random

00:08:30

and if it's less than 0.15 then I also

00:08:35

include this at the end of my prom

00:08:36

saying please include a random battle

00:08:38

encounter right so I think that's why

00:08:41

I'm getting attacked by goblins now is

00:08:43

because I added some difficulty to say

00:08:45

like add in some more encounters

00:08:48

um so now I'm fighting Undead creatures

00:08:50

and goblins so that kind of sucks anyway

00:08:52

I've been having a little bit of fun

00:08:53

with this just elevating my knowledge

00:08:54

with the open AI um I do plan to

00:08:56

hopefully make an actual interface for

00:08:58

this like a nexjs application UI that

00:09:02

kind of shows maybe images of the game

00:09:04

as we're playing and then also the

00:09:06

inventory I could use Dolly to generate

00:09:09

images for the icons that are coming in

00:09:11

I think that would be really cool as

00:09:13

well yeah that's all I got for you guys

00:09:15

have a good day and happy coding

Description:

📘 T3 Stack Tutorial: https://1017897100294.gumroad.com/l/jipjfm 🤖 SaaS I'm Building: https://icongeneratorai.com/ ✂️ Background Cutter: https://www.backgroundcutter.com/ 💬 Discord: https://discord.com/invite/4kGbBaa 🔔 Newsletter: https://webdevcody.com// 📁 GitHub: https://github.com/webdevcody 📺 Twitch: https://www.twitch.tv/webdevcody 🤖 Website: https://webdevcody.com 🐦 Twitter: https://twitter.com/webdevcody

Preparing download options

Popular

HD video

Only sound

All

* — If the video is playing in a new tab, go to it, then right-click on the video and select "Save video as..."

** — Link intended for online playback in specialized players

Questions about downloading video

How can I download "Why abstractions are useful (storing openai conversations into a db)" video?

http://unidownloader.com/ website is the best way to download a video or a separate audio track if you want to do without installing programs and extensions.
The UDL Helper extension is a convenient button that is seamlessly integrated into YouTube, Instagram and OK.ru sites for fast content download.
UDL Client program (for Windows) is the most powerful solution that supports more than 900 websites, social networks and video hosting sites, as well as any video quality that is available in the source.
UDL Lite is a really convenient way to access a website from your mobile device. With its help, you can easily download videos directly to your smartphone.

Which format of "Why abstractions are useful (storing openai conversations into a db)" video should I choose?

The best quality formats are FullHD (1080p), 2K (1440p), 4K (2160p) and 8K (4320p). The higher the resolution of your screen, the higher the video quality should be. However, there are other factors to consider: download speed, amount of free space, and device performance during playback.

Why does my computer freeze when loading a "Why abstractions are useful (storing openai conversations into a db)" video?

The browser/computer should not freeze completely! If this happens, please report it with a link to the video. Sometimes videos cannot be downloaded directly in a suitable format, so we have added the ability to convert the file to the desired format. In some cases, this process may actively use computer resources.

How can I download "Why abstractions are useful (storing openai conversations into a db)" video to my phone?

You can download a video to your smartphone using the website or the PWA application UDL Lite. It is also possible to send a download link via QR code using the UDL Helper extension.

How can I download an audio track (music) to MP3 "Why abstractions are useful (storing openai conversations into a db)"?

The most convenient way is to use the UDL Client program, which supports converting video to MP3 format. In some cases, MP3 can also be downloaded through the UDL Helper extension.

How can I save a frame from a video "Why abstractions are useful (storing openai conversations into a db)"?

This feature is available in the UDL Helper extension. Make sure that "Show the video snapshot button" is checked in the settings. A camera icon should appear in the lower right corner of the player to the left of the "Settings" icon. When you click on it, the current frame from the video will be saved to your computer in JPEG format.

What's the price of all this stuff?

It costs nothing. Our services are absolutely free for all users. There are no PRO subscriptions, no restrictions on the number or maximum length of downloaded videos.