This was a day-project during a hackathon to create an integration for Jira that utilizes an LLM somehow. This example was fully worked to the point of reading in an issue description and posting a comment on the initial Jira ticket.
Some requirements before we can start
- Node
- The
.nvmrcallows you to usenvm installandnvm useto use this project's supported node version.
- The
- AWS access
- The terminal the deploy command is run from
deploy-allrequires AWS auth. In my case this was done using Single Sign On and programatic access environment variable credentials.
- The terminal the deploy command is run from
- Don't forget to use
npm i
This project used some existant resources and cannot be deployed as is without modification. This project assumes a few things
- A lambda exists that handles prompt forwarding to an LLM generative AI service.
- This could be rectified by, instead of sending prompts to some assumed LLM forwarding lambda, using the Open AI API or some similar service's interface.
- My specific recommendation for a replacement would be to use this library as it fits well with this project and would require minimal LOE to implement: https://github.com/openai/openai-node
- A Jira auth token is saved to your
.envfile.- These can be a little tricky as there are many different versions of the Jira API t
- The one this project uses is v3
- These two Jira documentation points are a good starting point:
- These can be a little tricky as there are many different versions of the Jira API t
- This project is made fairly functionally, only using one singleton
JiraClient(which isn't strictly enforced). - There are no programatic tests written currently. This was tested in a non-production environment but was never considered for production.
- Tests could and should be added for the various pieces of this project if development was ever resumed.