Cloudwatch-bot

Set Up

1. Setting up Needed AWS Infrastrucure: Make sure that you have cloudtrail enabled and a Cloudwatch Log Group os created. You can use templates/logsinsight-bot.yaml to enable cloudtrail and create a loggroup
2. Sign up to OpenAI at https://openai.com and create an API Key

Tuning your own model

OpenAI allows you to tune your own versions of their model. for example you can tune the default davinci model on a specific use case

To do that:

1. set your envinroment with the following three parameters

export OPENAI_API_KEY=<Open AI Key>
export AWS_DEFAULT_REGION=<Region>
export LOGGROUP_NAME=<Cloud Trail LogGroup to search>

2. Install OpenAI tools

pip install openai

3. Create your sample tunning JSONL file with few hundreds examples Example:

{"prompt":"Show me the fields eventName, eventSource, source IP, with a limit of 79 ->","completion":" fields @timestamp, eventName, eventSource, sourceIPAddress | sort @timestamp desc | limit 79 ###"}

4. Create your tuned model

openai api fine_tunes.create -t <tuning-file>.jsonl -m davinci --suffix <Personal Identifier>

5. Set the value of FT_PERSONAL_MODEL to the newly created model name

export FT_PERSONAL_MODEL=<Model Name>

Use

Run cloudwatch-bot with a natural laguage question as a parameter

python3 bin/cloudwatch-bot.py <Questions>

Examples

python3 bin/cloudwatch-bot.py "show me the event name, ip address and arn with a limit of 7 ->"

Reference:

For more information visit OpenAi's documentation page https://beta.openai.com/docs/api-reference/introduction