|
| 1 | +# Introduction to Open AI |
| 2 | + |
| 3 | +## Overview |
| 4 | + |
| 5 | +What is [Open AI](https://openai.com/) ? |
| 6 | + |
| 7 | +* Research company on AI development |
| 8 | +* Builds and provides models |
| 9 | +* Builds and provides a standard protocol for using AI |
| 10 | + |
| 11 | +What is a model ? |
| 12 | + |
| 13 | +I see a model as a language super database. </br> |
| 14 | +Instead of writing a query, that is slow to query a traditional database like SQL, you can throw a question at a model and it gives you an answer really fast </br> |
| 15 | + |
| 16 | +Model examples: |
| 17 | +* GPT 3.5 |
| 18 | +* GPT 4 |
| 19 | + |
| 20 | +## Getting started |
| 21 | + |
| 22 | +The best way to get started and to understand OpenAI, is to learn hands on |
| 23 | + |
| 24 | +* Create an OpenAI account [here](https://openai.com/) |
| 25 | + |
| 26 | +## Chat GPT |
| 27 | + |
| 28 | +Here you can find the link to [ChatGPT](https://chat.openai.com/) |
| 29 | + |
| 30 | +## Open AI Playground |
| 31 | + |
| 32 | +Here you can find the link to the [OpenAI Playground](https://platform.openai.com/playground) |
| 33 | + |
| 34 | +## Build an AI powered app |
| 35 | + |
| 36 | +We can start with a `main.py` that reads a message |
| 37 | + |
| 38 | +``` |
| 39 | +import sys |
| 40 | +
|
| 41 | +message = sys.argv[0] |
| 42 | +
|
| 43 | +``` |
| 44 | +Then we will need the code from the Open AI playground and add it to our `main.py`. </br> |
| 45 | +Move the `import` statements to the top </br> |
| 46 | + |
| 47 | +Once you have tidied up everything, you can get the response message from the AI: |
| 48 | + |
| 49 | +``` |
| 50 | +responseMessage = response.choices[0].message.content |
| 51 | +``` |
| 52 | + |
| 53 | +Let's build our app |
| 54 | + |
| 55 | +``` |
| 56 | +cd ai\openai\introduction |
| 57 | +docker build . -t ai-app |
| 58 | +``` |
| 59 | + |
| 60 | +Set my OpenAI API key |
| 61 | + |
| 62 | +``` |
| 63 | +$ENV:OPENAI_API_KEY="" |
| 64 | +``` |
| 65 | + |
| 66 | +Run our AI App: |
| 67 | + |
| 68 | +``` |
| 69 | +docker run -it -e OPENAI_API_KEY=$ENV:OPENAI_API_KEY ai-app |
| 70 | +``` |
| 71 | + |
| 72 | +When we run the app, notice it has no concept of memory. </br> |
| 73 | +The playground works because it keeps track of all the user and AI messages and keeps appending new messages to it </br> |
| 74 | +So it can track the conversation. |
| 75 | + |
| 76 | +Let's keep track of messages, by writing it to a local file </br> |
| 77 | +We will also take the system message out and keep it as a constant in our code </br> |
| 78 | + |
| 79 | +Full example: |
| 80 | + |
| 81 | +``` |
| 82 | +import sys |
| 83 | +import os |
| 84 | +import json |
| 85 | +import openai |
| 86 | +
|
| 87 | +openai.api_key = os.getenv("OPENAI_API_KEY") |
| 88 | +
|
| 89 | +#read the incoming message |
| 90 | +message = sys.argv[1] |
| 91 | +user_message = { |
| 92 | + "role" : "user", |
| 93 | + "content" : message |
| 94 | +} |
| 95 | +
|
| 96 | +systemMessage = { |
| 97 | + "role": "system", |
| 98 | + "content": "You are a kubernetes exper that can assist developers with troubleshooting deployments\n\nTo help the developer you will need to know the namespaces as well as the pod name. Ask for missing information\n\nGenerate a command to help the developer surface logs or information\n" |
| 99 | +} |
| 100 | +
|
| 101 | +# read the cached user messages if there are any |
| 102 | +userMessages = [] |
| 103 | +if os.path.isfile("messages.json"): |
| 104 | + with open('messages.json', newline='') as messagesFile: |
| 105 | + data = messagesFile.read() |
| 106 | + userMessages = json.loads(data) |
| 107 | +
|
| 108 | +# add the new message to it and update the cached messages |
| 109 | +userMessages.append(user_message) |
| 110 | +with open('messages.json', 'w', newline='') as messagesFile: |
| 111 | + msgJSON = json.dumps(userMessages) |
| 112 | + messagesFile.write(msgJSON) |
| 113 | + print(msgJSON) |
| 114 | +
|
| 115 | +messages = [] |
| 116 | +messages.append(systemMessage) |
| 117 | +messages.extend(userMessages) |
| 118 | +
|
| 119 | +response = openai.ChatCompletion.create( |
| 120 | + model="gpt-3.5-turbo", |
| 121 | + messages=messages, |
| 122 | + temperature=1, |
| 123 | + max_tokens=256, |
| 124 | + top_p=1, |
| 125 | + frequency_penalty=0, |
| 126 | + presence_penalty=0 |
| 127 | +) |
| 128 | +
|
| 129 | +responseMessage = response.choices[0].message.content |
| 130 | +print(responseMessage) |
| 131 | +
|
| 132 | +``` |
| 133 | + |
| 134 | +Now we can mount our volume so we persist the cache of messages |
| 135 | + |
| 136 | +``` |
| 137 | +docker run -it -e OPENAI_API_KEY=$ENV:OPENAI_API_KEY -v ${PWD}:/app ai-app "can you help me with my deployment?" |
| 138 | +Of course! I'd be happy to help with your deployment. Could you please provide me with the namespace and the name of the pod you're encountering issues with? |
| 139 | +
|
| 140 | +docker run -it -e OPENAI_API_KEY=$ENV:OPENAI_API_KEY -v ${PWD}:/app ai-app "my pod is pod-123" |
| 141 | +Sure, I can help you with your deployment. Can you please provide me with the namespace in which the pod is running? |
| 142 | +
|
| 143 | +docker run -it -e OPENAI_API_KEY=$ENV:OPENAI_API_KEY -v ${PWD}:/app ai-app "its in the products namespace" |
| 144 | +Great! To surface the logs for the pod "pod-123" in the "products" namespace, you can use the following command: |
| 145 | +
|
| 146 | +```shell |
| 147 | +kubectl logs -n products pod-123 |
| 148 | +``` |
| 149 | + |
| 150 | +This command will retrieve the logs for the specified pod in the given namespace. Make sure you have the necessary permissions to access the namespace. |
| 151 | +``` |
0 commit comments