bentleymi / chatgpt-4-splunk Goto Github PK

Added cryptography and pcrypt but didnt use them

Need to pip uninstall --target=... cryptography pcrypt

Name of the app
ChatGPT 4 Splunk

Describe the bug
After having added the organization name and value and the API key and value, I am unable to perform a search using the openai command successfully.

To Reproduce
Steps to reproduce the behavior:

1. Install the plugin on Splunk.
2. In the "setup" page for the plugin

• enter the Organization Name in the "Enter Organization Name" field.
  ConductixWampfler
• enter the Organization Value in the "Enter Organization Value" field.
  org-blahblahblah
• enter the API Key Name in the "Enter API Key Name" field.
  CXW-Splunk
• enter the API Key Value in the "Enter API Key Value" field.
  sk-blahbalahasdlfkjasdf

3. Click on "Complete Setup"
4. From the "New Search" page/tab, add the following:
   host=namehere source!="Perfmon*" NOT "JIRA-STATS" sourcetype!="atlassian-jira-gc" | openai prompt="Given these logs, why did Jira stop responding?" model="gpt-3.5-turbo"
5. Limit the search to 15 minutes or some other small window.
6. Click on the magnifying glass icon.
7. Receive the following error (verbatim):
   command="openai", No such organization: org-ABCDEFG12345.

Expected behavior
I would expect:

1. the error display my accurate org key
2. To return results from OpenAI.

Screenshots

Splunk SOAR Version (please complete the following information):

• Unprivileged Install?
• running on a VM using Windows Server 2019 Datacenter, version 1809, OS build 17763.3770
• App Version: Splunk Enterprise 9.0.5 Build: e9494146ae5c

https://github.com/bentleymi/ta-openai-api/blob/258b8129d0167d12ce7f97406f8db2f63dd2fa40/TA-openai-api/bin/openai.py#L54

Line above references the curl command, should be changed to openai

Currently ChatGPT is not available by API. I originally thought the API was just earlier versions of the same thing but apparently ChatGPT is a specific version of the AI models that are not available via API. Therefore, in the next release we'll be renaming the app to TA-OpenAI-API and changing the command name to be '|openai' instead.

Looks like commands.conf is missing "enableheader=true"

The intended default was the latest but instead it uses version 2 instead of version 3.

chatgpt.py:92: engine = "text-davinci-002"

Should change to

chatgpt.py:92: engine = "text-davinci-003"

https://github.com/bentleymi/ta-openai-api/blob/b6644e771ef3588a33e2f1b12810e32c09c89824/TA-openai-api/bin/openai.py#L152

prompt=re.sub(r'[\n\r]+', '\n\n', result[options['field']])

Should be this instead:

prompt=re.sub(r'[\n\r]+', '\n\n', result[options['prompt_field']])

Use Case:

We allow multiple employees use ChatGPT at work, but they must use their work email account for their account.

It would be nice to add all of the users API information to Splunk so it would be easy to monitor what they are inputting to ensure PII or PHI information isn't being entering into OpenAPI Cloud Infrastructure.

If the user doesnt specify their orgid then the command will tell them to do so but gives them the old command syntax of "| chatgpt..."

https://github.com/bentleymi/ta-openai-api/blob/258b8129d0167d12ce7f97406f8db2f63dd2fa40/TA-openai-api/bin/openai.py#L72

Hi,

Thanks so much for this very useful application!

I have a situation, I have created a prompt that will ask the user to post the source code of a program written in Python, but when I put in the code, the system crashes as it is trying to integrate the code into the actual Simple XML code. I have enclosed the variable in quotes, used the encapsulation so it does not use the characters as Simple XML prompts, but still I having problems.

Here's an example of what I am trying to accomplish:

| openai prompt="Act as a Programmer with 10 years of experience writing code in Python and reviewing the code for bugs. I need you to review this code '$code.tok$' and see where the bugs are. Your output will be the analysis of the code and you will re-write it with your recommendations resolving the bugs found. Make sure you include a detailed analysis of your finding to the user learn for future projects."

So, the variable (token) will contain the actual source code in Python, but the prompt will see it as part of the Simple XML code and won't process it.

Any recommendation on how can I encapsulate the token, so it is treated as the text to be analyzed and does not break the program?

Your assistance is greatly appreciated.