Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions .vitepress/config.js
Original file line number Diff line number Diff line change
Expand Up @@ -38,6 +38,7 @@ const fullSidebar = [
},
{ text: "Core Concepts", link: "/getting-started/concepts" },
{ text: "CLI Commands", link: "/getting-started/cli" },
{ text: "Embedded Go API", link: "/getting-started/embedded" },
{ text: "AI Agent", link: "/getting-started/ai-agent" },
],
},
Expand Down
251 changes: 251 additions & 0 deletions getting-started/embedded.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,251 @@
# Embedded Go API

The Go package `github.com/dagucloud/dagu` exposes an experimental API for starting Dagu DAG runs from another Go application.

The API is marked experimental in the package documentation. Names, options, and behavior may change before this API is declared stable.

```go
import "github.com/dagucloud/dagu"
```

## Local Execution

`dagu.New` creates an embedded engine backed by Dagu's file stores. `RunFile` and `RunYAML` load a DAG, start the run asynchronously, and return a `*dagu.Run`.

```go
package main

import (
"context"
"fmt"
"log"
"time"

"github.com/dagucloud/dagu"
)

func main() {
ctx, cancel := context.WithTimeout(context.Background(), time.Minute)
defer cancel()

engine, err := dagu.New(ctx, dagu.Options{
HomeDir: "/var/lib/myapp/dagu",
})
if err != nil {
log.Fatal(err)
}
defer func() {
if err := engine.Close(context.Background()); err != nil {
log.Printf("close Dagu engine: %v", err)
}
}()

run, err := engine.RunYAML(ctx, []byte(`
name: embedded-local
params:
- MESSAGE
steps:
- name: hello
command: echo "${MESSAGE}"
`), dagu.WithParams(map[string]string{
"MESSAGE": "hello from the host app",
}))
if err != nil {
log.Fatal(err)
}

status, err := run.Wait(ctx)
if err != nil {
log.Fatal(err)
}
fmt.Printf("%s finished with %s\n", status.RunID, status.Status)
}
```
Comment on lines +15 to +63
Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

⚠️ Potential issue | 🔴 Critical

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
# Description: Verify dagu package API matches documentation examples

# Check if the package exists and extract key API signatures
rg -A 10 'func New\(' --type go
rg -A 5 'func.*RunYAML' --type go
rg -A 5 'func.*WithParams' --type go
rg -A 10 'type Status struct' --type go

Repository: dagucloud/docs

Length of output: 653


🏁 Script executed:

#!/bin/bash
# Explore the repository structure to understand what's available
find . -type f -name "*.go" | head -20
git ls-files | head -30
ls -la

Repository: dagucloud/docs

Length of output: 2217


🌐 Web query:

dagucloud/dagu package API documentation go

💡 Result:

The dagucloud/dagu package refers to the Go module for Dagu, a lightweight workflow orchestration engine. Its API documentation is available on pkg.go.dev at https://pkg.go.dev/github.com/dagu-org/dagu (latest v1.30.3 as of 2026-01-04). Note that there appear to be multiple related repositories (dagu-org/dagu, dagucloud/dagu, yohamta/dagu), but dagu-org/dagu is the primary active one with the most recent pkg.go.dev page. This page provides standard Go package documentation including: - Module overview, license (GPL-3.0), Go version compatibility (1.25+). - Dependencies list. - README with installation, quick start, features (YAML workflows, Web UI, CLI, distributed execution via gRPC, AI agent, etc.). - Links to full docs at docs.dagu.sh, GitHub examples, and community. Dagu also exposes a REST API (documented at https://docs.dagu.sh/web-ui/api and https://docs.dagu.cloud/reference/api) for HTTP-based interaction (e.g., /api/v1 or /api/v2 endpoints for DAG management, health checks), but this is the server's HTTP API, not the Go package API. No separate Go client SDK is explicitly documented in results; use the core package for embedding or direct use. For programmatic control, interact via the REST API or gRPC (mentioned for distributed mode). Source code at https://github.com/dagucloud/dagu or https://github.com/dagu-org/dagu.

Citations:


🌐 Web query:

site:pkg.go.dev github.com/dagu-org/dagu New RunYAML WithParams

💡 Result:

No explicit documentation or source code snippets for a "RunYAML WithParams" function or method in github.com/dagu-org/dagu were found in the search results. The pkg.go.dev page for github.com/dagu-org/dagu v1.30.3 (most recent, published 2026-01-04) provides general project information, installation, CLI usage like "dagu start hello.yaml", and YAML workflow examples with params in sub-DAG calls (e.g., params: "SOURCE=s3://bucket/data.csv"), but no Go API details matching the query. The official docs are referenced at docs.dagu.sh, but no specific Go library API for RunYAML is detailed in results. The CLI "dagu start <yaml>" likely handles YAML execution, possibly with params via flags or env vars, but no "WithParams" identified. Repository likely at https://github.com/dagu-org/dagu or https://github.com/dagucloud/dagu based on installer scripts.

Citations:


🌐 Web query:

github.com/dagu-org/dagu embedding API Go example

💡 Result:

No public example exists for embedding the Dagu workflow engine API directly in Go code. Dagu (github.com/dagu-org/dagu) is primarily a standalone binary workflow engine using YAML-defined DAGs, with REST API for programmatic control when the server runs (docs.dagu.sh/overview/api). pkg.go.dev shows the module exists but no documented public embedding API or examples. An open GitHub issue (#583) requests adding direct Go execution like dagu.Start("dagName", dagu.Params("key=val")), but it remains unresolved as of latest updates. To use Dagu programmatically, run the binary via exec.Command or interact via its gRPC/REST interfaces in distributed mode. For embedding workflows in Go, consider alternatives like github.com/rhosocial/go-dag.

Citations:


🌐 Web query:

site:github.com dagucloud/dagu "func New" "RunYAML"

💡 Result:

No function named "func New" referencing or using "RunYAML" was found in the dagucloud/dagu repository or its documentation across multiple searches targeting Go source code, GitHub files, and official docs. The repository is a Go-based workflow engine (https://github.com/dagucloud/dagu) where workflows are defined in YAML and executed via CLI commands like dagu start hello.yaml or API endpoints that accept inline YAML specs (e.g., POST /api/v2/dag-runs with "spec" field). There is an open issue (#583) proposing a public Go API for direct DAG execution like dagu.Start("dagName"), but it remains unimplemented as of 2026-04-19. Code snippets mentioning YAML parsing exist (e.g., migration to goccy/go-yaml in #913), but no matching "func New" with "RunYAML". Searches for source files like internal/digraph/loader.go surfaced coverage reports but no explicit code with the terms. Official docs at docs.dagu.sh focus on YAML workflows and CLI/API usage, with no embedded Go library API documented.

Citations:


🌐 Web query:

dagu-org/dagu Go API embedded example "dagu.New"

💡 Result:

No complete, up-to-date embedded Go API example using dagu.New was found in official documentation or pkg.go.dev pages. Dagu (github.com/dagu-org/dagu) is primarily designed as a standalone binary for YAML-based workflow execution via CLI commands like dagu start workflow.yaml, with a web UI and HTTP/gRPC APIs for management. The pkg.go.dev page (v1.30.3) shows the module but no public Go API documentation for direct embedding like dagu.New. An open GitHub issue #583 (dated 2024-06-03, updated 2025-12-06, still open) requests exactly this feature: "Enable Direct Execution of DAGs via Go Code" with proposed API like dagu.Start("someDAG", dagu.Params("FOO=BAR")). This confirms the feature is planned but not implemented as of the latest sources (repo active as of 2026-04). For embedding workflows in Go applications, current alternatives are: - Execute the dagu binary as a subprocess using os/exec (common for CLI tools). - Use the HTTP/gRPC client APIs once server is running (dagu start-all exposes /api/v1). Example subprocess usage: package main import ( "os/exec" ) func main { cmd := exec.Command("dagu", "start", "hello.yaml") err := cmd.Run if err != nil { panic(err) } } Official docs at docs.dagu.sh focus on YAML/CLI/server usage, not programmatic Go embedding. Repo: https://github.com/dagu-org/dagu (3K+ stars, Go primary lang).

Citations:


Remove or replace this example—the documented API does not exist in the actual dagucloud/dagu package.

The example shows dagu.New(), engine.RunYAML(), run.Wait(), and dagu.WithParams() functions that are not currently implemented in the github.com/dagu-org/dagu package. The package is primarily a CLI tool and server with REST/gRPC APIs, not a Go library for direct programmatic embedding. An open GitHub issue (#583, still unresolved) requests exactly this feature.

Either remove this example, replace it with an actual working approach (e.g., executing the dagu binary via os/exec or using the REST API when the server runs), or clearly mark it as aspirational/future API subject to change.

🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.

In `@getting-started/embedded.md` around lines 15 - 63, The embedded Go example
must be removed or replaced because the functions dagu.New, engine.RunYAML,
run.Wait, and dagu.WithParams do not exist in the actual dagu package; either
delete this example or replace it with a working pattern (e.g., demonstrate
invoking the dagu CLI via os/exec or calling the server REST/gRPC API) and
update the text to explicitly state that a native embedding API is not yet
available (referencing issue `#583`) if you keep an aspirational sample. Ensure
references to dagu.New, engine.RunYAML, run.Wait, and dagu.WithParams are
eliminated or clearly marked as future/placeholder APIs.


The same engine can load a workflow file:

```go
run, err := engine.RunFile(ctx, "/var/lib/myapp/workflows/report.yaml")
```

`Wait` blocks until the run reaches a terminal state or the context is canceled. It returns a non-nil status when the final status can be read. A non-success terminal status is returned as an error.

## Engine Options

`dagu.Options` maps to Dagu's existing config and file-backed storage paths.

| Field | Effect |
| --- | --- |
| `HomeDir` | Dagu application home used for default config and data paths. |
| `ConfigFile` | Explicit Dagu config file. |
| `DAGsDir` | Directory used to resolve named DAGs and sub-DAGs. |
| `DataDir` | File-backed state directory. |
| `LogDir` | Run log directory. |
| `ArtifactDir` | Artifact directory. |
| `BaseConfig` | Base config file applied while loading DAGs. |
| `Logger` | `*slog.Logger` used by the embedded engine. A quiet logger is used when nil. |
| `DefaultMode` | Default execution mode for runs that do not pass `WithMode`. |
| `Distributed` | Coordinator, TLS, worker selector, and status polling options for distributed mode. |

`Close` releases engine-owned resources. It does not replace `Run.Wait`, `Run.Stop`, or worker shutdown.

## Run Options

Run options apply to a single `RunFile` or `RunYAML` call.

| Option | Effect |
| --- | --- |
| `WithRunID(id)` | Sets an explicit DAG run ID. |
| `WithName(name)` | Overrides the loaded DAG name. |
| `WithParams(map[string]string)` | Sets DAG parameters from a Go map. |
| `WithParamsList([]string)` | Sets DAG parameters from `KEY=VALUE` entries. |
| `WithDefaultWorkingDir(dir)` | Sets the default working directory while loading the DAG. |
| `WithMode(mode)` | Overrides the engine default execution mode. |
| `WithWorkerSelector(map[string]string)` | Sets the distributed worker selector for one run. |
| `WithTags(tags...)` | Adds tags to one run. |
| `WithDryRun(enabled)` | Enables or disables dry-run mode. |

## Run Status And Cancellation

The run handle exposes the run reference and current state:

```go
ref := run.Ref()
status, err := run.Status(ctx)
err = run.Stop(ctx)
```

`Engine.Status(ctx, ref)` and `Engine.Stop(ctx, ref)` operate on a run reference. `Engine.Status` reads the local file-backed run status. For distributed runs, use the `Run` returned by `RunFile` or `RunYAML`; its `Status` method reads status from the coordinator.

`dagu.Status` contains the DAG name, run ID, attempt ID, status string, start and finish times, error string, log file path, archive directory, worker ID, and trigger type.

## Custom Executors

`RegisterExecutor` registers a process-local executor type that can be used in DAG YAML.

Registration mutates process-global state. Register custom executors before concurrent DAG loading or execution. `UnregisterExecutor` exists for tests and should not run concurrently with engine use.

```go
dagu.RegisterExecutor(
"embedded_echo",
func(_ context.Context, step dagu.Step) (dagu.Executor, error) {
return &echoExecutor{step: step}, nil
},
dagu.WithExecutorCapabilities(dagu.ExecutorCapabilities{Command: true}),
)
```

The executor type name must start with a letter and may contain letters, digits, `_`, and `-`.

```go
type echoExecutor struct {
step dagu.Step
stdout io.Writer
}

func (e *echoExecutor) SetStdout(out io.Writer) {
e.stdout = out
}

func (e *echoExecutor) SetStderr(io.Writer) {}

func (e *echoExecutor) Kill(os.Signal) error {
return nil
}

func (e *echoExecutor) Run(context.Context) error {
out := e.stdout
if out == nil {
out = io.Discard
}
_, err := fmt.Fprintf(out, "handled %s\n", e.step.Name)
return err
}
```

YAML can then reference the registered type:

```yaml
name: embedded-custom
steps:
- name: call-go-code
type: embedded_echo
command: domain operation from DAG YAML
```

A custom executor is registered only in the current Go process. In distributed mode, every worker process that may execute a DAG containing that type must register the same executor type before it starts processing tasks.

## Distributed Execution

Embedded distributed execution dispatches a loaded DAG to an existing Dagu coordinator. The public embedded API does not start a coordinator; start one with Dagu server commands such as `dagu coord` or `dagu start-all` with coordinator settings.

The embedded coordinator client requires TLS configuration unless plaintext is explicitly enabled. For a local plaintext coordinator:

```go
TLS: dagu.TLSOptions{Insecure: true}
```

For TLS connections, use `CertFile`, `KeyFile`, `ClientCAFile`, and optionally `SkipTLSVerify`.

```go
engine, err := dagu.New(ctx, dagu.Options{
HomeDir: "/var/lib/myapp/dagu",
DefaultMode: dagu.ExecutionModeDistributed,
Distributed: &dagu.DistributedOptions{
Coordinators: []string{"127.0.0.1:50055"},
TLS: dagu.TLSOptions{Insecure: true},
WorkerSelector: map[string]string{
"pool": "embedded",
},
PollInterval: time.Second,
},
})
```

`RunFile` and `RunYAML` load and validate the DAG in the caller process, then dispatch the loaded YAML definition to the coordinator. Shared-nothing workers receive the DAG definition from the coordinator and do not read the caller's DAG file path.

An embedded worker can run in the same host process:

```go
worker, err := engine.NewWorker(dagu.WorkerOptions{
ID: "embedded-worker-1",
MaxActiveRuns: 4,
Labels: map[string]string{"pool": "embedded"},
})
if err != nil {
return err
}

workerCtx, stopWorker := context.WithCancel(ctx)
defer stopWorker()

go func() {
if err := worker.Start(workerCtx); err != nil && !errors.Is(err, context.Canceled) {
log.Printf("worker stopped: %v", err)
}
}()

if err := worker.WaitReady(ctx); err != nil {
return err
}

run, err := engine.RunFile(ctx, "/var/lib/myapp/workflows/report.yaml")
if err != nil {
return err
}
status, err := run.Wait(ctx)
```

`WorkerOptions.Coordinators` overrides `DistributedOptions.Coordinators` when non-empty. `WorkerOptions.TLS` overrides `DistributedOptions.TLS` when non-zero. `HealthPort` starts the worker health endpoint on that port; `0` disables it.

## Repository Examples

The Dagu repository contains runnable examples:

```sh
go run ./examples/embedded/local
go run ./examples/embedded/custom-executor
DAGU_COORDINATORS=127.0.0.1:50055 go run ./examples/embedded/distributed
```

The distributed example expects a coordinator to already be listening at the address in `DAGU_COORDINATORS`.