Axis and Allies GPT

A & A Gee Pee Tee solves the problem of "I want to play A&A but I don't want to read the rulebook". Solution: spin up an "AI Assistant" with a custom domain of knowledge concerning the board game.

• we discussed this when playing the game
• today it came out
• read [[1]]
• the data is here: https://axisallies.com/rules/axis-allies-rules-1942-2nd-edition.pdf

Work Notes

• Installed Python locally sans admin access https://plainenglish.io/blog/install-python-on-a-locked-down-pc-without-local-admin-37a440c42c12
• Created API Key https://platform.openai.com/api-keys
• added API key to environment variable
• run sample code. doesn't work.
• reboot for my environment variable to "stick"
• error 429 exceeded quota... find out that my account is out of credits or it expired because it is > 3 months old https://stackoverflow.com/questions/75898276/openai-chatgpt-gpt-3-5-api-error-429-you-exceeded-your-current-quota-please
• upgrade/buy credits https://platform.openai.com/account/billing/overview added $50 bucks.
• recreate API key
• reboot? nah, just had to refresh my path by closing command prompt and re-opening. tested by

echo %OPENAI_API_KEY%

• re-run

python openai-test.py

SUCCESS.

okay, now on to XY...

• https://platform.openai.com/docs/quickstart?context=python
• https://platform.openai.com/docs/assistants/how-it-works/runs-and-run-steps
• https://platform.openai.com/docs/overview

The Manual

the PDF is kinda big... (47000KB). Converted it to text (100KB) with an online tool. i uploaded (read) it using

# Read the contents of the text file (user manual)
file_path = "axy.txt"  # Replace with your file path
with open(file_path, "r", encoding="utf-8") as file:
   user_manual_text = file.read()

and then fought the syntax here for way too long:

# Define a function to interact with ChatGPT
def ask_chatgpt(question, context):
response = openai.completions.create(
  engine="text-davinci-002",
   prompt=f"User Manual: {context}\nQuestion: {question}\nAnswer:",
    max_tokens=150,  # Adjust based on your needs
     temperature=0.7,  # Adjust for creativity vs. accuracy
  )
   return response.choices[0].text.strip()

The rest of the code seemed straightworward:

# Query ChatGPT with your question
question = "What is the general combat sequence"
response = ask_chatgpt(question, user_manual_text)

response = client.completions.create(
  model="gpt-3.5-turbo-instruct",
  prompt="What is the general combat sequence"
)

#Print the response
print(f"Question: {question}")
print(f"Answer: {response}")

but sadly the syntax for

 response = openai.completions.create( ... 

is pretty unclear to me at this point and even chatGPT was tripping up with it and apologizing.

Roll your own assistant

So, I ended up rolling my own assistant in the playground and tweeted it here:

https://twitter.com/phalseid/status/1722022484927447146