feat: add agentSchema and agentCode sync actions to ComponentBase

[ZdenekSrotyr]

Summary

Adds two new built-in sync actions to ComponentBase that enable KAI (Keboola AI) to introspect and execute code against configured components at runtime:

• agentSchema: Auto-generates a schema describing the component via inspect — includes source code of the component class and imported modules, registered sync actions with docstrings, context variables with their types/methods, and installed packages. Components can override get_agent_schema() to customize or return None to disable.
• agentCode: Executes arbitrary Python code (from parameters.agent_code) inside the component's container. The code namespace includes comp (the initialized component instance) plus anything from get_agent_context(). Results are returned via a result variable set in the executed code.

Component authors only need to override get_agent_context() (e.g., return {"sf": authenticated_client}) to expose initialized clients. Everything else is handled by the base class.

Review & Testing Checklist for Human

• Security of exec() call: agent_code runs exec(code, {'__builtins__': __builtins__}, local_ns) — full Python builtins are available, including __import__, open, eval. Confirm that the container sandbox is sufficient and that this action cannot be triggered outside the expected isolated Docker context. This is the highest-risk item.
• _generate_agent_schema calls get_agent_context() during schema generation (line 330): If a component's get_agent_context() initializes an authenticated client, agentSchema will attempt authentication just to populate context_variables. Decide whether schema generation should work without live credentials.
• Global action registration: agentSchema and agentCode are registered in _SYNC_ACTION_MAPPING for ALL ComponentBase subclasses at import time. Verify this doesn't conflict with existing components and that Developer Portal registration is the actual gating mechanism.
• Non-serializable results from agentCode: If executed code sets result to a non-JSON-serializable object, process_sync_action_result will raise. No guard exists for this — confirm whether that's acceptable or if a serialization fallback is needed.
• End-to-end test plan: Register agentSchema + agentCode actions for a real component (e.g., component-salesforce-v2) in Developer Portal, run agentSchema to verify the generated schema is useful, then run agentCode with a real snippet against live credentials to verify sub-second execution.

Notes

• Tests mutate config.json files on disk with try/finally restore — functional but fragile if test infrastructure changes.
• inspect.getsource() failures are silently caught, so schema may be sparse for C-extension-heavy components.
• Link to Devin run: https://app.devin.ai/sessions/dda90ce482014efa8ba751869407fdd4
• Requested by: @ZdenekSrotyr

Release Notes

Justification, description

Adds agentSchema and agentCode sync actions to enable KAI to programmatically introspect and interact with configured components at runtime with sub-second latency.

Plans for Customer Communication

N/A — internal capability for KAI integration

Impact Analysis

Low risk to existing components — new actions are registered but require Developer Portal registration to be exposed. Components can opt-out by overriding get_agent_schema() to return None.

Deployment Plan

Standard release process

Rollback Plan

Revert commit and release previous version

Post-Release Support Plan

N/A

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring