Go for GenAI!

Keynote presented at DevFest Bletchley Park 2025 highlighting the new developments in Go for GenAI

1. Bletchley Park Daniela Petruzalek Developer Relations, Google Cloud Go for GenAI!
2. About me… DevRel at Google UK Originally from Brazil Backend / Data Engineer Currently obsessed with AI Love Games, Anime and Cats =^_^=
3. Agenda • Why Go? • Genkit Go • Agent Development Kit • Go SDK for MCP
4. Why Go?
5. Why Go? Because it has a cute mascot! Bletchley Park
6. Or because… It’s a compiled language Code is simple to read / maintain It has a rich standard library Great performance and tooling It is fun!* * based on my totally biased personal opinion
7. The two sides of GenAI Code generation (aka “vibe coding”) AI workloads in production
8. Go for GenAI: Code generation Compiled: tight feedback loop Code is standardised: better for training models Standard library: less moving parts Tooling: automate code quality checks
9. Go for GenAI: Production Workloads Great performance Concurrency out-of-the-box (go func()) Tooling (go vet, trace, pprof, static analysis …) Rich ecosystem of AI frameworks
10. Genkit is an open-source framework for building full-stack AI-powered applications Bletchley Park
11. Key Features Model support: Google, Ollama, OpenAI, … Building blocks: structured output, tool calling, multi-modal input/output and more. Developer tools: CLI and developer UI Production monitoring: Firebase dashboard
12. // Genkit: Define and Run a Flow g := genkit.Init(ctx, genkit.WithPlugins(&googlegenai.VertexAI{}) ) helloFlow := genkit.DefineFlow(g, "helloFlow", func(ctx context.Context, prompt string) (string, error) { resp, _ := genkit.Generate(ctx, g, ai.WithModelName("vertexai/gemini-2.5-flash"), ai.WithPrompt(prompt)) return resp.Text(), nil }, ) result, _ := helloFlow.Run(ctx, "Say hello!") fmt.Println(result)
13. // Genkit: Define a Flow Server g := genkit.Init(ctx, genkit.WithPlugins(&googlegenai.VertexAI{}) ) genkit.DefineFlow(g, "helloFlow", func(ctx context.Context, prompt string) (string, error) { resp, _ := genkit.Generate(ctx, g, ai.WithModelName("vertexai/gemini-2.5-flash"), ai.WithPrompt(prompt)) return resp.Text(), nil }, ) // Keep running to serve the flow <-ctx.Done()
15. Agent Development Kit (ADK) for Go An open-source toolkit for building, evaluating, and deploying AI agents Bletchley Park
16. Key Features Model agnostic, deployment agnostic Agentic workflows: sequential, parallel, loop, … Pre-built (search, code exec) and custom tools Built-in evaluation
17. // ADK: Create and Launch an Agent model, _ := gemini.NewModel(ctx, "gemini-2.5-flash", &genai.ClientConfig{Backend: genai.BackendVertexAI}) agent, _ := llmagent.New(llmagent.Config{ Name: "hello_agent", Model: model, Instruction: "You are a helpful assistant.", }) full.NewLauncher().Execute(ctx, &adk.Config{ AgentLoader: services.NewSingleAgentLoader(agent), }, []string{"console"})
19. Genkit vs ADK
21. Model Context Protocol (MCP) is an open standard that allows AI models to connect and communicate with external tools and data sources Bletchley Park Go SDK for MCP
22. Shameless plug… Session: Hello, MCP World! @ 12:20 PM Lab: Build an MCP with Gemini CLI and Go @ 3:45 PM
23. Bletchley Park Daniela Petruzalek Developer Relations, Google Cloud danicat.dev | @danicat83 Thank you!