bee-hive in kubernetes

[akihikokuroda]

This PR makes Bee-hive deployable and accessible in kubernetes.

A small API is added bee-hive image. It accepts request with 2 yaml files and runs workflow.
This PR has kubernetes deployment and service (it's using a NodePort right now) yaml files. The can be deployed with the kubectl. The bee-hive with container runtime is also adjusted to use the API instead of mounting the local file system.

The bee-hive.sh has 4 commands (the first argument)

• build: build container image
• deploy: deploy bee-hive in container service
• deploy-k: deploy bee-hive in kubernetes
• run: run the work flow in the environment

Run workflow in local environment with container runtime (e.g. podman) or kubernetes cluster

• Prepare required environments

  • Run a local instance of the Bee Stack
  • curl is required for run command
  • for Kubernetes, kubectl and sed are required

• Build bee-hive container image: ./bee-hive.sh build in bee-hive/bee-hive directory

• Deploy bee-hive: ./bee-hive.sh deploy BEE_API=http://xxx.xxx.xxx.xxx:4000 BEE_API_KEY=sk-proj-testkey

  • To deploy to kubernetes cluster, user deploy-k instead deploy as the first argument
  • The required environment variables can be provided at the end of command arguments (e.g. BEE_API=http://192.168.86.45:4000).

• Prepare agent and workflow definition yaml files in the current directory (e.g. agent.yaml, workflow.yaml)

• Run workflow: ./bee-hive.sh run agents.yaml workflow.yaml

[psschwei]

Local deploy works as expected.
Checking Kubernetes deploy-k now.

[akihikokuroda]

@psschwei They are using same local port number. You may need to stop the docker container first.

[psschwei]

Kubernetes is failing for me with an "unable to create agents" error. Partial stack trace

Traceback (most recent call last):
  File "/usr/src/app/api.py", line 49, in process_workflow
    workflow_instance = Workflow(agents_yaml, workflow_yaml[0])
                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/src/app/bee_hive/workflow.py", line 47, in __init__
    self.agents[agent_def["metadata"]["name"]] = get_agent_class(
                                                 ^^^^^^^^^^^^^^^^
  File "/usr/src/app/bee_hive/bee_agent.py", line 43, in __init__
    assistant = client.beta.assistants.create(
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/site-packages/openai/resources/beta/assistants.py", line 149, in create
    return self._post(
           ^^^^^^^^^^^
  File "/usr/local/lib/python3.11/site-packages/openai/_base_client.py", line 1283, in post
    return cast(ResponseT, self.request(cast_to, opts, stream=stream, stream_cls=stream_cls))
                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/site-packages/openai/_base_client.py", line 960, in request
    return self._request(
           ^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/site-packages/openai/_base_client.py", line 1005, in _request
    return self._retry_request(
           ^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/site-packages/openai/_base_client.py", line 1098, in _retry_request
    return self._request(
           ^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/site-packages/openai/_base_client.py", line 1005, in _request
    return self._retry_request(
           ^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/site-packages/openai/_base_client.py", line 1098, in _retry_request
    return self._request(
           ^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/site-packages/openai/_base_client.py", line 1015, in _request
    raise APITimeoutError(request=request) from err
openai.APITimeoutError: Request timed out.

[psschwei]

I wonder if the problem is my Kubernetes not being able to talk to the bee-stack 🤔 ... (I'm using kind)

[akihikokuroda]

@psschwei Does the deployment get the correct environment variables?

[maximilien]

Thank you so much for this @akihikokuroda and will spend some time on this but don't let me slow you down

@psschwei is reviewing and if he reaches LGTM before I get to review then push forward and merge and I will catch up. This is v1 of a very important feature.

どうもありがとうございます

[akihikokuroda]

@psschwei I tried with kind and it is working. I changed the bee-hive.yaml at the last minutes. You might not pick up the change.

[psschwei]

awesome! I was able to run without issue.

I have one minor comment on the readme (which could be fixed now or in a followon PR), but otherwise LGTM

[psschwei]

nit: let's make two sections for the deploy, one for docker and one for kubernetes. that would also allow for including some kubernetes-specific instructions:

• to load the image into the cluster
• to port-forward the bee-hive service, if necessary

since both of those will differ depending on the k8s used (rancher, kind, etc.), I don't think we need to include specific instructions, but we should mention that the steps are needed

[akihikokuroda]

@psschwei Thanks! I'll create another PR for the readme update!

[psschwei]

feel free to merge when you're ready (or if you want to wait to see if anyone has comments that's cool too)