Enable the use of self-hosted LLMs

[artivis]

This PR introduces several fixes and features that enable the use of other, open-ai api compatible, LLMs such as those served by the Ollama framework.

Most notably:

• Make use of the api endpoint
• Add a --dry-run option to exec that prints the answer without executing it
• Enable setting up the system prompts through env var & cli arg

[fujitatomoya]

@artivis thanks 👍 you are the 1st contributor 🥇 i will definitely look at the fix.

[fujitatomoya]

So this actually is a part of #21.

what this is trying to do is to be agnostic from some OpenAI default and required setting so that https://ollama.com/ can be configured and used as backend AI system.

for doing this, IMO we can take this fix to support but there are things need to be done.

• Documentation update. all of the docs are currently related to OpenAI, documentation needs to be more generic and supported AI backend system.
• Refactor code base to be agnostic from OpenAI only. https://ollama.com/ can be a good example for this.

[fujitatomoya]

@artivis i added the minor patch to fix the problem with verification.sh script. after building, verification for all container images, i will take this in.