Struct RawTask 

pub struct RawTask {

    pub id: Spanned<String>,
    pub description: Option<Spanned<String>>,
    pub action: Option<RawTaskAction>,
    pub provider: Option<Spanned<String>>,
    pub model: Option<Spanned<String>>,
    pub with_refs: Option<Spanned<IndexMap<Spanned<String>, Spanned<String>>>>,
    pub depends_on: Option<Spanned<Vec<Spanned<String>>>>,
    pub output: Option<Spanned<RawOutputConfig>>,
    pub for_each: Option<Spanned<RawForEach>>,
    pub retry: Option<Spanned<RawRetryConfig>>,
    pub decompose: Option<Spanned<DecomposeSpec>>,
    pub concurrency: Option<Spanned<u32>>,
    pub fail_fast: Option<Spanned<bool>>,
    pub structured: Option<StructuredOutputSpec>,
    pub artifact: Option<Spanned<Value>>,
    pub log: Option<Spanned<Value>>,
    pub span: Span,
}

A raw task as parsed from YAML.

All string references (task IDs, aliases) are unresolved. Resolution happens in Phase 2 (analyzed AST).

Fields

id: Spanned<String>

Task identifier (must be unique within workflow)

description: Option<Spanned<String>>

Human-readable description

action: Option<RawTaskAction>

The action this task performs (infer, exec, fetch, invoke, agent)

provider: Option<Spanned<String>>

Task-specific provider override

model: Option<Spanned<String>>

Task-specific model override

with_refs: Option<Spanned<IndexMap<Spanned<String>, Spanned<String>>>>

Binding declarations: with: { alias: "expression" }

Values are raw strings that get parsed by parse_with_entry() in Phase 2. Examples:

• data: step1 — simple task reference
• temp: step1.data.temp ?? 20 — path + default
• cfg: $env.API_KEY — environment binding
• val: step1.output | upper | trim — with transforms

depends_on: Option<Spanned<Vec<Spanned<String>>>>

Explicit ordering dependencies: depends_on: [task_id1, task_id2]

These are pure ordering edges — no data flows through them. Data dependencies are expressed via with: bindings.

output: Option<Spanned<RawOutputConfig>>

Output configuration

for_each: Option<Spanned<RawForEach>>

For-each iteration

retry: Option<Spanned<RawRetryConfig>>

Retry configuration

decompose: Option<Spanned<DecomposeSpec>>

Decompose modifier for runtime DAG expansion

concurrency: Option<Spanned<u32>>

Standalone concurrency (used with decompose when no for_each)

fail_fast: Option<Spanned<bool>>

Standalone fail_fast (used with decompose when no for_each)

structured: Option<StructuredOutputSpec>

Structured output specification (JSON schema enforcement)

artifact: Option<Spanned<Value>>

Artifact output configuration (persists task output to files)

log: Option<Spanned<Value>>

Log configuration override for this task

span: Span

The span of the entire task block

Implementations

impl RawTask

pub fn new(id: impl Into<String>) -> RawTask

Create a new raw task with the given ID.

pub fn has_dependencies(&self) -> bool

Check if this task has any dependencies (with: or depends_on:).

pub fn depends_on_ids(&self) -> Vec<&str>

Get all explicit depends_on task IDs.

Trait Implementations

impl Clone for RawTask

fn clone(&self) -> RawTask

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for RawTask

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.

impl Default for RawTask

fn default() -> RawTask

Returns the “default value” for a type.