feat(all): add support for ADK

[EItanya]

This is the official PR to add support for ADK to kagent 🎉

This PR modifies many core parts of the system:

1. kagent-engine has been removed
2. Each agent now gets its own deployment which runs ADK
3. An SDK will be released to allow users to BYO their own ADK agents

[Copilot]

Pull Request Overview

This PR adds support for ADK (Agent Development Kit) to kagent, representing a significant architectural shift from the previous kagent-engine approach to individual agent deployments running ADK. Key changes include removal of the centralized engine, introduction of ADK integration with new Python packages, and updates to the Kubernetes deployment configuration.

• Complete removal of kagent-engine and related autogen integration
• Introduction of new kagent-adk Python package for ADK agent integration
• Migration from Python 3.12 to 3.13 and restructured workspace configuration

Reviewed Changes

Copilot reviewed 148 out of 219 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
python/src/autogenstudio/	Removed entire autogenstudio module including web managers, validation, and evaluation components
python/packages/kagent-adk/	New ADK integration package with agent configuration models and A2A protocol support
python/packages/kagent/	Restructured kagent package with simplified CLI for static agent deployment
python/pyproject.toml	Converted to uv workspace configuration with updated dependencies
helm/kagent/	Updated Helm chart removing engine deployment and adding grafana-mcp sidecar
go/internal/	Removed autogen client integration and simplified HTTP server handlers

Comments suppressed due to low confidence (1)

python/packages/kagent-adk/src/kagent_adk/_session_service.py:17

• [nitpick] The class name KAgentSessionService is inconsistent with the filename which uses lowercase 'kagent'. Consider renaming to KagentSessionService to match the original naming convention or update the filename.

class KAgentSessionService(BaseSessionService):

[Copilot]

The Gemini model type handling is incomplete. When self.model.type == "gemini", the code assigns model = self.model.model which appears to be a string, but the Agent constructor expects a model instance like the other branches. This will likely cause a runtime error.

Suggested change

	model = self.model.model
	model = GeminiLLM(model=self.model.model)

[Copilot]

The file is opened in text mode without specifying encoding. Consider adding encoding='utf-8' to ensure consistent behavior across different platforms and avoid potential encoding issues.

Suggested change

	with open(filepath, "r") as f:
	with open(filepath, "r", encoding="utf-8") as f:

[Copilot]

Using the latest tag for the mcp/grafana image in production is a security and stability risk. The image should be pinned to a specific version to ensure reproducible deployments and avoid unexpected changes.

Suggested change

	image: mcp/grafana:latest
	image: mcp/grafana:8.5.0

[Copilot]

[nitpick] The commented out caching logic # self._runner = resolved_runner suggests intended functionality that was disabled. Consider either implementing proper caching or removing the comment to avoid confusion about the intended behavior.

Suggested change

	# self._runner = resolved_runner
	self._cached_runner = resolved_runner

[Copilot]

The Grafana API key is being set as a plain environment variable. For better security, consider using a Kubernetes Secret and referencing it via valueFrom.secretKeyRef instead of exposing the API key in the deployment manifest.

Suggested change

	value: {{ .Values.tools.grafana.apiKey | quote }}
	valueFrom:
	secretKeyRef:
	name: grafana-api-key-secret
	key: apiKey

[yuval-k]

approving based off of previous PR comments

[nashtsai]

hi, I am curious that is there any reason for switching from Autogen (https://kagent.dev/docs/kagent/concepts/architecture has yet updated) to ADK?