# mk-agents-python-sdk-tutorial

let's make this a very basic tutorial on the agents api

## steps

### basics

- create python notebook in google colab
- get all your secret stuff together and add it to secrets. The bare minimum will be a value for `OPENAI_API_KEY`.
- Mount Google Drive so you can read and write files.

from google.colab import drive
drive.mount('/content/drive')

- get and store your secret values (only include the ones you want to use)

from google.colab import userdata
OPENAI_API_KEY = userdata.get('OPENAI_API_KEY')
SLACK_WEBHOOK_URL = userdata.get('SLACK_WEBHOOK_URL')
AIRTABLE_API_KEY = userdata.get('AI_LAB_AIRTABLE_API_KEY')
AIRTABLE_BASE_ID = userdata.get('AI_LAB_AIRTABLE_BASE_ID')

- if you want to define an input or output folder or path, do that

output_folder="/content/drive/MyDrive/ll_color_colabs/_projects/_ai-lab-planning/openai_agents_sdk_tutorials/output"

- install the openai and openai-agents packages

!pip install openai, openai-agents

- do your first basic openai call to test and you should see a response

from openai import OpenAI
client = OpenAI(api_key=OPENAI_API_KEY)

response = client.responses.create(
    model="gpt-4o",
    input="Write a one-sentence bedtime story about a unicorn."
)

print(response.output_text)

- investigate the content of the response in different ways, first to get json:

print(response.model_dump_json(indent=2))

- and then to see the "help" docs (which is how you would have found out about `response.output_text` if you didn't know it was there)

help(response)

### agents

For traces to work (a way of tracking what you're doing) we've found you need to do this (well, at least the second):

from agents import set_default_openai_key, set_tracing_export_api_key

set_default_openai_key(OPENAI_API_KEY)
set_tracing_export_api_key(OPENAI_API_KEY)

## sidequests

- [file search](https://platform.openai.com/docs/guides/tools-file-search)
- [vector stores](https://platform.openai.com/docs/guides/embeddings)
- [notes on retrieval](https://platform.openai.com/docs/guides/retrieval)
- [conversation state](https://platform.openai.com/docs/guides/conversation-state?api-mode=responses)
-

## links and to-dos

- [OpenAI Agents SDK Docs on the github.io page](https://openai.github.io/openai-agents-python/)
- and the [OpenAI Agents SDK Docs on the OpenAI API docs page](https://platform.openai.com/docs/guides/agents-sdk). `#TODO: at some point we'll want to check to see if these are different`
-
- direct [link to md doc in the repo that describes agent ensemble handoff patterns](https://github.com/openai/openai-agents-python/blob/main/examples/agent_patterns/README.md). And here is a version for us to copy paste or add to in hackmd: [openai-common-agentic-patterns](/U_BSNbHNQ4uuA2tNb-Km4g)
- dev
- [mk-working-folder](https://drive.google.com/drive/folders/16ct5o29ic9jywIXa6DQZcAiJB0rwDy8C?usp=drive_link)
- [openai builder lab repo](https://github.com/openai/openai-builder-lab/tree/main) seems like the monorepo to imitate more or less in tutorials?
- https://github.com/openai/openai-realtime-console
-

## knowledge base

- [openai-openapi-schema](https://github.com/openai/openai-openapi/blob/master/openapi.yaml)

## questions

- should we also involve langchain and langraph? got this far in a useful video:
<iframe width="560" height="315" src="https://www.youtube.com/embed/aHCDrAbH_go?si=IZPJ__jcPnTgmRx2&amp;start=410" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" referrerpolicy="strict-origin-when-cross-origin" allowfullscreen></iframe>

### langGraph comparisons

https://langchain-ai.github.io/langgraph/tutorials/introduction/