Struct Task 

pub struct Task {

    pub id: String,
    pub with_spec: Option<WithSpec>,
    pub output: Option<OutputPolicy>,
    pub decompose: Option<DecomposeSpec>,
    pub for_each: Option<Value>,
    pub for_each_as: Option<String>,
    pub concurrency: Option<usize>,
    pub fail_fast: Option<bool>,
    pub action: TaskAction,
    pub artifact: Option<ArtifactSpec>,
    pub log: Option<LogConfig>,
    pub depends_on: Option<Vec<String>>,
    pub structured: Option<StructuredOutputSpec>,
}

Fields

id: String

with_spec: Option<WithSpec>

Typed binding system

with: block for binding task outputs to local aliases.

Example

tasks:
  - id: process
    with:
      summary: $step1.abstract | lower | trim
      count: $step1.items | length ?? 0
    infer: "Process: {{with.summary}} ({{with.count}} items)"

output: Option<OutputPolicy>

Output format and validation

decompose: Option<DecomposeSpec>

Runtime DAG expansion via semantic traversal

When specified, the task will be decomposed at runtime based on graph traversal results. This takes precedence over static for_each.

Example

tasks:
  - id: generate_children
    decompose:
      strategy: semantic
      traverse: HAS_CHILD
      source: $entity
    infer: "Generate for {{with.item}}"

for_each: Option<Value>

Parallel iteration over array values

When specified, the task will be executed once for each value in the array. Each iteration runs in parallel with its own bindings.

Example

tasks:
  - id: process_locales
    for_each: ["en-US", "fr-FR", "de-DE"]
    as: locale
    exec:
      command: "echo {{with.locale}}"

for_each_as: Option<String>

Variable name for current iteration value

Defaults to “item” if not specified. The value is accessible as {{with.<as>}} in templates.

concurrency: Option<usize>

Maximum parallel executions for for_each

Controls how many iterations run concurrently. Defaults to 1 (sequential). Set higher for parallel execution.

Example

for_each: ["a", "b", "c", "d", "e"]
concurrency: 3  # Run at most 3 at a time

fail_fast: Option<bool>

Stop all iterations on first error

When true (default), aborts remaining iterations if any fails. When false, continues executing remaining iterations.

Example

for_each: $items
fail_fast: false  # Continue even if some fail

action: TaskAction

artifact: Option<ArtifactSpec>

Artifact output configuration for this task

Can be a simple boolean to enable/disable, a single output spec, or an array of output specs for multiple artifacts.

log: Option<LogConfig>

Task-level logging configuration

Overrides workflow-level log settings for this task.

depends_on: Option<Vec<String>>

Explicit task dependencies

Task IDs that must complete before this task can execute.

Accepts both flow and depends_on as field names.

Example

- id: process
  depends_on: [fetch_data, validate]
  infer: "Process {{with.data}}"

structured: Option<StructuredOutputSpec>

Structured output configuration

When specified, enforces JSON Schema validation on task output. Uses the 4-layer StructuredOutputEngine for ~99.99% compliance.

Example

- id: extract_data
  infer: "Extract user data"
  structured: ./schemas/user.json

Or with full configuration:

- id: extract_data
  infer: "Extract user data"
  structured:
    schema: ./schemas/user.json
    max_retries: 3
    enable_repair: true

Implementations

impl Task

pub fn validate_for_each(&self) -> Result<(), NikaError>

Validate for_each configuration

Returns error if:

• for_each is not an array and not a binding expression
• for_each array is empty

Binding expressions (strings containing {{) are accepted because they resolve to arrays at runtime.

pub fn has_for_each(&self) -> bool

Check if this task has for_each iteration

pub fn for_each_var(&self) -> &str

Get the iteration variable name (defaults to “item”)

pub fn for_each_concurrency(&self) -> usize

Get the concurrency limit for for_each (defaults to 1 = sequential)

pub fn for_each_fail_fast(&self) -> bool

Get the fail_fast setting for for_each (defaults to true)

pub fn has_decompose(&self) -> bool

Check if this task has decompose modifier

pub fn decompose_spec(&self) -> Option<&DecomposeSpec>

Get the decompose spec if present

pub fn action_icon(&self) -> &'static str

Get the action icon for TUI display

Returns an emoji icon based on the task’s verb type. Canonical icons from CLAUDE.md:

• ⚡ infer (LLM generation)
• 📟 exec (Shell command)
• 🛰️ fetch (HTTP request)
• 🔌 invoke (MCP tool)
• 🐔 agent (Agentic loop - parent)
• 🐤 subagent (spawned via spawn_agent)

pub fn subagent_icon() -> &'static str

Get the icon for a subagent (spawned via spawn_agent)

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

Get list of task IDs this task depends on

Returns task IDs from the depends_on field.

Trait Implementations

impl Clone for Task

fn clone(&self) -> Task

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Task

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for Task

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.