Flyte enables production-grade orchestration for machine learning workflows and data processing created to accelerate local workflows to production.

Flyte

I'm working through setting up a flyte agent at the moment. I find it a bit rough around the edges but the latest issue I've had is that the docker image that is built <https://docs.flyte.org/en/latest/flyte_agents/developing_agents.html#build-a-new-docker-image|per the docs> blows up with complains about missing `prometheus_client`. It seems like this should come as a dependency of <https://github.com/flyteorg/flytekit/blob/master/flytekit/extend/backend/agent_service.py#L26|flytekit> no? I don't see it in the obvious <https://github.com/flyteorg/flytekit/blob/master/pyproject.toml|requirements files>.

Oh. Looks like the repo has a good example of what the agent dockerfile should actually look like: <https://github.com/flyteorg/flytekit/blob/76fb7c344162b7fae141c40c2fc8d18a71091fc2/Dockerfile.agent#L10>

thanks. we didn’t add it to flytekit since it only used by agent. Maybe we should add the dependency to extra.
like `pip install flytekit[agent]`

Have a couple PRs, more to come.
<https://github.com/flyteorg/flyte/pull/5398|https://github.com/flyteorg/flyte/pull/5398>
• Markdown syntax error. Separated out because I don't think this should require any back and forth. 
<https://github.com/flyteorg/flyte/pull/5399|https://github.com/flyteorg/flyte/pull/5399>
• Improve `developing_agents` and `local_python_environment` docs with complete running examples of an example agent task
• This dumbs down the task a little bit but I think thats better than a non functional skeleton. 
• Adds additional comments around the agent workflow and any gotchas (matching agent/task type, ensuring agent/task registration so pyflyte doesn't just noop with zero errors)