Genkit Go
Genkit Go is an AI SDK for Go that provides generation, structured output,
streaming, tool calling, prompts, and flows with a unified interface across
model providers.
Hello World
package main
import (
"context"
"fmt"
"log"
"net/http"
"github.com/firebase/genkit/go/ai"
"github.com/firebase/genkit/go/genkit"
"github.com/firebase/genkit/go/plugins/googlegenai"
"github.com/firebase/genkit/go/plugins/server"
)
func main() {
ctx := context.Background()
g := genkit.Init(ctx, genkit.WithPlugins(&googlegenai.GoogleAI{}))
genkit.DefineFlow(g, "jokeFlow", func(ctx context.Context, topic string) (string, error) {
return genkit.GenerateText(ctx, g,
ai.WithModelName("googleai/gemini-flash-latest"),
ai.WithPrompt("Tell me a joke about %s", topic),
)
})
mux := http.NewServeMux()
for _, f := range genkit.ListFlows(g) {
mux.HandleFunc("POST /"+f.Name(), genkit.Handler(f))
}
log.Fatal(server.Start(ctx, "127.0.0.1:8080", mux))
}
Core Features
Load the appropriate reference based on what you need:
| Feature |
Reference |
When to load |
| Initialization |
references/getting-started.md |
Setting up genkit.Init, plugins, the *Genkit instance pattern |
| Generation |
references/generation.md |
Generate, GenerateText, GenerateData, streaming, output formats, Embed |
| Prompts |
references/prompts.md |
DefinePrompt, DefineDataPrompt, .prompt files, schemas |
| Tools |
references/tools.md |
DefineTool, DefineResumableTool interrupts (Interrupted claim, Restart/Respond, WithResume), ai/tool helpers (tool.Interrupt, tool.Fail, tool.AttachParts) |
| Middleware |
references/middleware.md |
ai.Middleware, ai.WithUse, Hooks (Generate/Model/Tool), built-ins (Retry, Fallback, ToolApproval, Filesystem, Skills, SoftToolErrors, ContextCompression, Budget) |
| Observability |
references/observability.md |
OpenTelemetry export (otelgenai), tracing.SetInstrumentation, aix.Observer, reflection API outside dev |
| Flows & HTTP |
references/flows-and-http.md |
DefineFlow, DefineStreamingFlow, genkit.Handler, HTTP serving |
| Model Providers |
references/providers.md |
Google AI, Vertex AI, Anthropic, OpenAI-compatible, Ollama setup |
Agents (Experimental)
Genkit Go has an experimental agent API for persistent, multi-turn
conversations (sessions, snapshots, interrupts, branching, background
execution).
It is gated: initialize with genkit.Init(ctx, genkit.WithExperimental()) or
the
constructors panic. Server constructors come from genkit/exp (aliased
genkitx); types and options from ai/exp (aliased aix); session stores from
ai/exp/localstore.
- Agent or flow? If the task is conversational, multi-turn, or described as
"an agent", "assistant", or "chatbot", build it with
genkitx.DefineAgent
rather than hand-rolling a Generate + tools loop in a flow. Reach for a
plain flow only for single-shot, stateless generation.
For details see:
- Agents: defining/serving an agent, running turns,
and client- vs server-managed state (start here).
- Sessions & persistence: session stores
(
localstore.NewInMemorySessionStore/NewFileSessionStore, PostgreSQL),
snapshots, and the sessionstoretest conformance suite.
- Human-in-the-loop / interrupts:
pausing for approval/input and resuming.
- Branching: forking a conversation from a
snapshot.
- Background agents: detaching long-running
turns and polling.
- Working with state: typed custom session
state, streamed as JSON patches.
- Artifacts: producing and reading named
deliverables.
- Multi-agent orchestration: delegating to
sub-agents, background delegation,
continue_task, and remote agents
(DefineRemoteAgent).
- Advanced custom agents:
DefineCustomAgent
for full turn control.
- Deploying agents: serving agents over
HTTP with
genkit.Handler.
Generative UI (A2UI)
Genkit Go has an A2UI (Agent-to-UI) plugin
(github.com/firebase/genkit/go/plugins/a2ui/exp, imported as a2uix) that
lets an agent stream interactive UI surfaces (cards, lists, forms, buttons),
not just prose. The whole integration is the &a2uix.Surfaces{} model
middleware added via ai.WithUse
on a generate call or agent inline prompt. Go is server-only: it has no A2UI
client/renderer. Render surfaces with a JS or Dart/Flutter client that talks
to
your Go agent over HTTP (the server and clients are wire-compatible).
- A2UI: server middleware, options, custom catalogs, and
security. For the client (rendering surfaces), refer to the A2UI reference
in the Genkit JS or Dart skill.
Genkit CLI (recommended)
genkit start unintrusively wraps any Go program that uses the Genkit library,
running it unchanged while capturing traces from every Genkit action so you can
prove tools were actually called and inspect model I/O from the terminal, even
for headless checks. It forwards stdio, so interactive CLI tools that rely on
stdin/stdout work without issues. Running the app directly (go run .) skips
trace capture, so you're debugging blind. Check install with genkit --version.
Installation:
curl -sL cli.genkit.dev | bash
Primary pattern (default): prefix genkit start -- to your normal run
command. This collects telemetry from any Genkit code your program runs, whether
triggered from the dev UI, your own web server/web UI, or a plain script. Starts
the Developer UI (usually http://localhost:4000) for running flows, model and
agent playground, and browsing traces:
genkit start -- go run .
genkit start --noui -- go run . # same, without the Dev UI (still a persistent server)
genkit start -o -- go run . # also opens the browser
genkit start runs until you stop it with Ctrl+C. That is expected and correct
for the common cases: a server your web/mobile app calls, or an interactive CLI
you exit yourself. --noui only drops the Dev UI; it is not a one-shot
command and will not exit on its own. Do not use genkit start as a
blocking step in automated/non-interactive contexts; use flow:run (below) for
that.
Non-interactive use (agents/CI): add the global --non-interactive flag
before -- so the CLI uses defaults and never blocks on a prompt (e.g. the
first-run analytics notice): genkit start --non-interactive -- go run . (works
with flow:run too).
Run a flow (flow:run): invoke a specific flow by name from the CLI. Append
your run command after -- to spin up the runtime just for this run (the
command runs as-is to register your flows):
genkit flow:run myFlow '{"data": "input"}' -- go run .
genkit flow:run myFlow '{"data": "input"}' --stream -- go run . # with streaming
genkit flow:run myFlow '{"data": "input"}' --wait -- go run . # wait for completion
This is self-terminating: it runs the flow once, prints a Trace ID, then
exits, so it's the right choice for a quick, non-interactive check (unlike
genkit start). Traces for this run can be inspected using the trace commands
below.
Debugging with traces: the fastest way to see prompts, model inputs/outputs,
tool calls, latencies, and errors. Inspect from the terminal after any run under
genkit start:
genkit trace:list # find recent trace IDs
genkit trace:get <traceId> # full trace details (inputs, outputs, tool calls, errors)
genkit trace:get <traceId> --format json # machine-readable JSON, safe to pipe into jq or other parsers
For machine-readable output, pass --format json to get clean JSON you can pipe
into jq or other parsers. The default output is human-oriented (banner/log
lines, possible truncation on large traces), so don't pipe that form directly;
use --format json, grep, or the Dev UI trace viewer.
Documentation:
genkit docs:search "streaming" go
genkit docs:list go
genkit docs:read go/flows.md
See references/getting-started.md for full CLI
and Developer UI details.
Key Guidance
- Pass
g explicitly. The *Genkit instance returned by genkit.Init is
the central registry. Pass it to all Genkit functions rather than storing it
as a global. This is a core pattern throughout the SDK.
- Wrap AI logic in flows. Flows give you tracing, observability, HTTP
deployment via
genkit.Handler, and the ability to test from the Developer UI
and CLI. Any generation call worth keeping should live in a flow.
- Verify with traces, not a blind run. Running the app directly (
go run .)
does not capture dev traces. See the Genkit CLI
section for how to run your app and capture traces.
- Use
jsonschema:"description=..." struct tags on output types. The model
uses these descriptions to understand what each field should contain. Without
them, structured output quality drops significantly.
- Write good tool descriptions. The model decides which tools to call based
on their description string. Vague descriptions lead to missed or incorrect
tool calls.
- Use
.prompt files for complex prompts. They separate prompt content from
Go code, support Handlebars templating, and can be iterated on without
recompilation. Code-defined prompts are better for simple, single-line cases.
- Use
DefineResumableTool for interrupts. Claim paused calls with the
tool's Interrupted method and resume with ai.WithResume.
ai.InterruptWith, RestartWith/RespondWith, and WithToolRestarts are
deprecated; genkitx.DefineInterruptibleTool and ai/exp/tool were removed
in v1.14.0. See references/tools.md.
- Reach for built-in middleware before writing one.
Retry, Fallback,
ToolApproval, Filesystem, Skills, and SoftToolErrors cover the common
cross-cutting needs and compose with each other via ai.WithUse. See
references/middleware.md. When you do write custom
middleware, allocate per-call state in closures captured by New, and guard
anything that WrapTool mutates because tools may run concurrently.
- Look up the latest model IDs. Model names change frequently. Check
provider documentation for current model IDs rather than relying on hardcoded
names. See references/providers.md.
1---2name: developing-genkit-go3description: >-4 Provides guidance on developing AI-powered applications using Genkit in Go.5 Use when the user asks to build AI features, agents, flows, or tools in Go6 using Genkit, or when working with Genkit Go code involving generation,7 prompts, streaming, tool calling, or model providers. Don't use for Genkit in8 JavaScript/TypeScript, Python, or Dart.9metadata:10 source: https://skillmd.com/skills/google/developing-genkit-go11 version: "1.0.0"12 category: AiAndMachineLearning13---1415# Genkit Go1617Genkit Go is an AI SDK for Go that provides generation, structured output,18streaming, tool calling, prompts, and flows with a unified interface across19model providers.2021## Hello World2223```go24package main2526import (27 "context"28 "fmt"29 "log"30 "net/http"3132 "github.com/firebase/genkit/go/ai"33 "github.com/firebase/genkit/go/genkit"34 "github.com/firebase/genkit/go/plugins/googlegenai"35 "github.com/firebase/genkit/go/plugins/server"36)3738func main() {39 ctx := context.Background()40 g := genkit.Init(ctx, genkit.WithPlugins(&googlegenai.GoogleAI{}))4142 genkit.DefineFlow(g, "jokeFlow", func(ctx context.Context, topic string) (string, error) {43 return genkit.GenerateText(ctx, g,44 ai.WithModelName("googleai/gemini-flash-latest"),45 ai.WithPrompt("Tell me a joke about %s", topic),46 )47 })4849 mux := http.NewServeMux()50 for _, f := range genkit.ListFlows(g) {51 mux.HandleFunc("POST /"+f.Name(), genkit.Handler(f))52 }53 log.Fatal(server.Start(ctx, "127.0.0.1:8080", mux))54}55```5657## Core Features5859Load the appropriate reference based on what you need:6061| Feature | Reference | When to load |62| --- | --- | --- |63| Initialization | [references/getting-started.md](references/getting-started.md) | Setting up `genkit.Init`, plugins, the `*Genkit` instance pattern |64| Generation | [references/generation.md](references/generation.md) | `Generate`, `GenerateText`, `GenerateData`, streaming, output formats, `Embed` |65| Prompts | [references/prompts.md](references/prompts.md) | `DefinePrompt`, `DefineDataPrompt`, `.prompt` files, schemas |66| Tools | [references/tools.md](references/tools.md) | `DefineTool`, `DefineResumableTool` interrupts (`Interrupted` claim, `Restart`/`Respond`, `WithResume`), `ai/tool` helpers (`tool.Interrupt`, `tool.Fail`, `tool.AttachParts`) |67| Middleware | [references/middleware.md](references/middleware.md) | `ai.Middleware`, `ai.WithUse`, `Hooks` (Generate/Model/Tool), built-ins (`Retry`, `Fallback`, `ToolApproval`, `Filesystem`, `Skills`, `SoftToolErrors`, `ContextCompression`, `Budget`) |68| Observability | [references/observability.md](references/observability.md) | OpenTelemetry export (`otelgenai`), `tracing.SetInstrumentation`, `aix.Observer`, reflection API outside dev |69| Flows & HTTP | [references/flows-and-http.md](references/flows-and-http.md) | `DefineFlow`, `DefineStreamingFlow`, `genkit.Handler`, HTTP serving |70| Model Providers | [references/providers.md](references/providers.md) | Google AI, Vertex AI, Anthropic, OpenAI-compatible, Ollama setup |7172## Agents (Experimental)7374Genkit Go has an **experimental** agent API for persistent, multi-turn75conversations (sessions, snapshots, interrupts, branching, background76execution).77It is gated: initialize with `genkit.Init(ctx, genkit.WithExperimental())` or78the79constructors panic. Server constructors come from `genkit/exp` (aliased80`genkitx`); types and options from `ai/exp` (aliased `aix`); session stores from81`ai/exp/localstore`.8283- **Agent or flow?** If the task is conversational, multi-turn, or described as84 "an agent", "assistant", or "chatbot", build it with `genkitx.DefineAgent`85 rather than hand-rolling a `Generate` + tools loop in a flow. Reach for a86 plain flow only for single-shot, stateless generation.8788For details see:8990- [Agents](references/agents.md): defining/serving an agent, running turns,91 and client- vs server-managed state (start here).92- [Sessions & persistence](references/agents-sessions.md): session stores93 (`localstore.NewInMemorySessionStore`/`NewFileSessionStore`, PostgreSQL),94 snapshots, and the `sessionstoretest` conformance suite.95- [Human-in-the-loop / interrupts](references/agents-human-in-the-loop.md):96 pausing for approval/input and resuming.97- [Branching](references/agents-branching.md): forking a conversation from a98 snapshot.99- [Background agents](references/agents-background.md): detaching long-running100 turns and polling.101- [Working with state](references/agents-state.md): typed custom session102 state, streamed as JSON patches.103- [Artifacts](references/agents-artifacts.md): producing and reading named104 deliverables.105- [Multi-agent orchestration](references/agents-multi-agent.md): delegating to106 sub-agents, background delegation, `continue_task`, and remote agents107 (`DefineRemoteAgent`).108- [Advanced custom agents](references/agents-custom.md): `DefineCustomAgent`109 for full turn control.110- [Deploying agents](references/agents-deployment.md): serving agents over111 HTTP with `genkit.Handler`.112113## Generative UI (A2UI)114115Genkit Go has an **A2UI** (Agent-to-UI) plugin116(`github.com/firebase/genkit/go/plugins/a2ui/exp`, imported as `a2uix`) that117lets an agent stream interactive UI **surfaces** (cards, lists, forms, buttons),118not just prose. The whole integration is the `&a2uix.Surfaces{}` model119middleware added via `ai.WithUse`120on a generate call or agent inline prompt. **Go is server-only: it has no A2UI121client/renderer.** Render surfaces with a JS or Dart/Flutter client that talks122to123your Go agent over HTTP (the server and clients are wire-compatible).124125- [A2UI](references/a2ui.md): server middleware, options, custom catalogs, and126 security. For the client (rendering surfaces), refer to the A2UI reference127 in the Genkit JS or Dart skill.128129## Genkit CLI (recommended)130131`genkit start` unintrusively wraps any Go program that uses the Genkit library,132running it unchanged while capturing traces from every Genkit action so you can133prove tools were actually called and inspect model I/O from the terminal, even134for headless checks. It forwards stdio, so interactive CLI tools that rely on135stdin/stdout work without issues. Running the app directly (`go run .`) skips136trace capture, so you're debugging blind. Check install with `genkit --version`.137138**Installation:**139```bash140curl -sL cli.genkit.dev | bash141```142143**Primary pattern (default):** prefix `genkit start --` to your normal run144command. This collects telemetry from any Genkit code your program runs, whether145triggered from the dev UI, your own web server/web UI, or a plain script. Starts146the Developer UI (usually http://localhost:4000) for running flows, model and147agent playground, and browsing traces:148```bash149genkit start -- go run .150genkit start --noui -- go run . # same, without the Dev UI (still a persistent server)151genkit start -o -- go run . # also opens the browser152```153`genkit start` runs until you stop it with Ctrl+C. That is expected and correct154for the common cases: a server your web/mobile app calls, or an interactive CLI155you exit yourself. `--noui` only drops the Dev UI; it is **not** a one-shot156command and will not exit on its own. Do **not** use `genkit start` as a157blocking step in automated/non-interactive contexts; use `flow:run` (below) for158that.159160**Non-interactive use (agents/CI):** add the global `--non-interactive` flag161before `--` so the CLI uses defaults and never blocks on a prompt (e.g. the162first-run analytics notice): `genkit start --non-interactive -- go run .` (works163with `flow:run` too).164165**Run a flow (`flow:run`):** invoke a specific flow by name from the CLI. Append166your run command after `--` to spin up the runtime just for this run (the167command runs as-is to register your flows):168```bash169genkit flow:run myFlow '{"data": "input"}' -- go run .170genkit flow:run myFlow '{"data": "input"}' --stream -- go run . # with streaming171genkit flow:run myFlow '{"data": "input"}' --wait -- go run . # wait for completion172```173This is **self-terminating**: it runs the flow once, prints a `Trace ID`, then174exits, so it's the right choice for a quick, non-interactive check (unlike175`genkit start`). Traces for this run can be inspected using the trace commands176below.177178**Debugging with traces:** the fastest way to see prompts, model inputs/outputs,179tool calls, latencies, and errors. Inspect from the terminal after any run under180`genkit start`:181```bash182genkit trace:list # find recent trace IDs183genkit trace:get <traceId> # full trace details (inputs, outputs, tool calls, errors)184genkit trace:get <traceId> --format json # machine-readable JSON, safe to pipe into jq or other parsers185```186187For machine-readable output, pass `--format json` to get clean JSON you can pipe188into `jq` or other parsers. The **default** output is human-oriented (banner/log189lines, possible truncation on large traces), so don't pipe that form directly;190use `--format json`, grep, or the Dev UI trace viewer.191192**Documentation:**193```bash194genkit docs:search "streaming" go195genkit docs:list go196genkit docs:read go/flows.md197```198199See [references/getting-started.md](references/getting-started.md) for full CLI200and Developer UI details.201202## Key Guidance203204- **Pass `g` explicitly.** The `*Genkit` instance returned by `genkit.Init` is205 the central registry. Pass it to all Genkit functions rather than storing it206 as a global. This is a core pattern throughout the SDK.207- **Wrap AI logic in flows.** Flows give you tracing, observability, HTTP208 deployment via `genkit.Handler`, and the ability to test from the Developer UI209 and CLI. Any generation call worth keeping should live in a flow.210- **Verify with traces, not a blind run.** Running the app directly (`go run .`)211 does not capture dev traces. See the [Genkit CLI](#genkit-cli-recommended)212 section for how to run your app and capture traces.213- **Use `jsonschema:"description=..."` struct tags on output types.** The model214 uses these descriptions to understand what each field should contain. Without215 them, structured output quality drops significantly.216- **Write good tool descriptions.** The model decides which tools to call based217 on their description string. Vague descriptions lead to missed or incorrect218 tool calls.219- **Use `.prompt` files for complex prompts.** They separate prompt content from220 Go code, support Handlebars templating, and can be iterated on without221 recompilation. Code-defined prompts are better for simple, single-line cases.222- **Use `DefineResumableTool` for interrupts.** Claim paused calls with the223 tool's `Interrupted` method and resume with `ai.WithResume`.224 `ai.InterruptWith`, `RestartWith`/`RespondWith`, and `WithToolRestarts` are225 deprecated; `genkitx.DefineInterruptibleTool` and `ai/exp/tool` were removed226 in v1.14.0. See [references/tools.md](references/tools.md).227- **Reach for built-in middleware before writing one.** `Retry`, `Fallback`,228 `ToolApproval`, `Filesystem`, `Skills`, and `SoftToolErrors` cover the common229 cross-cutting needs and compose with each other via `ai.WithUse`. See230 [references/middleware.md](references/middleware.md). When you do write custom231 middleware, allocate per-call state in closures captured by `New`, and guard232 anything that `WrapTool` mutates because tools may run concurrently.233- **Look up the latest model IDs.** Model names change frequently. Check234 provider documentation for current model IDs rather than relying on hardcoded235 names. See [references/providers.md](references/providers.md).