feat: Session 3 - Core Semantic Kernel Port - Basic Interfaces

[devin-ai-integration]

Description

This PR implements Session 3 of the Go migration: Core Semantic Kernel Port - Basic Interfaces. It establishes the foundational Go interfaces and types that will serve as the backbone for the agent framework, translating the Python semantic kernel concepts into idiomatic Go patterns.

⚠️ Important: This session focuses on interface definitions only. All LLM provider implementations return NotImplementedError placeholders and will be completed in future sessions.

Changes

Core Kernel Interfaces (pkg/kernel/):

• ChatHistory and ChatMessageContent types for conversation management
• ContentItem interface with TextContent and ImageContent implementations
• FunctionCallContent and FunctionResultContent for tool calling
• Kernel and KernelFunction interfaces for orchestration
• AuthorRole enum (user, assistant, system, tool)

Agent Builder Pattern (pkg/agents/):

• AgentBuilder with fluent configuration interface
• SKAgent wrapper type with model metadata
• ChatCompletionAgent interface for agent implementations
• AgentConfig YAML configuration structure
• basicChatAgent stub (returns "not implemented" - intentional)

Model Provider Abstraction (pkg/models/):

• Provider interface for LLM client abstraction
• ChatCompletionFactory for provider management
• Stub implementations for OpenAI, Anthropic, and Google providers
• Model capability detection (structured output support)
• Provider-specific error types

Plugin System (pkg/plugins/):

• Plugin and KernelFunction interfaces
• PluginFunction implementation with parameter specs
• ExtraDataCollector for metadata collection
• PluginLoader interface for registration/retrieval

Error Handling (pkg/errors/):

• Framework-specific error types with Go error wrapping
• AuthenticationError, AgentInvokeError, PersistenceError
• HITLInterventionRequired for human-in-the-loop flows

Documentation:

• Updated GO_MIGRATION.md to mark Sessions 2 and 3 complete

Type of Change

• New feature
• Bugfix
• Refactor
• Documentation
• Other

Key Review Points

Interface Design:

• Do the Go interfaces properly capture the Python semantic kernel patterns?
• Are the type relationships and dependencies correct?
• Does the builder pattern provide sufficient flexibility for agent configuration?

Go Idioms:

• Proper use of context.Context for cancellation
• Error handling with wrapping (fmt.Errorf with %w)
• Channel-based streaming instead of async iterables
• Interface definitions in consumer packages

Placeholder Implementations:

• All model providers return NotImplementedError{Feature: "X client creation"} - this is intentional
• basicChatAgent.Invoke() and InvokeStream() return "not implemented yet" - this is intentional
• These will be completed in Session 4 (Sequential Agent) and Session 5 (Chat Agent)

Integration Points:

• Dependencies on existing kernel.ModelType and kernel.ChatCompletionClient from previous sessions
• YAML configuration compatibility with existing Python format

Additional Comments

This session establishes the structural foundation for the Go migration. The interfaces are designed to support:

• Multiple LLM providers (OpenAI, Anthropic, Google)
• Function/tool calling with structured parameters
• Streaming responses via Go channels
• Human-in-the-loop approval workflows
• Plugin-based extensibility

Next Steps: Session 4 will implement the Sequential agent execution logic using these interfaces.

Link to Devin run: https://app.devin.ai/sessions/8073faecb180437987bd162e6e9d07c3
Requested by: Jude Partovi (@jpartovi)

[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

[jpartovi]

Devin, can you remove the updates from the README?