Skip to main content
T1 starter · every office job — the first example with infer.schema: — the model doesn’t get to answer in prose. It returns the JSON shape you declared, or the engine rejects it.

​
The job

The meeting ended an hour ago and nobody remembers who owns what. The transcript is sitting in a file. This workflow extracts every action item as {owner, task, due} — typed, validated, importable by whatever tracker you use.

​
The shape

​
The file

t1-meeting-actions.nika.yaml
nika: v1
workflow: meeting-actions
description: "Transcript → typed action items {owner, task, due}"

model: mock/echo            # swap for openai/gpt-5.2 or any provider in the catalog

vars:
  transcript_path:
    type: string
    required: true
    description: "Path to the raw meeting transcript"

tasks:
  - id: transcript
    invoke:
      tool: "nika:read"
      args: { path: "${{ vars.transcript_path }}" }

  - id: extract
    depends_on: [transcript]
    infer:
      prompt: |
        Extract every action item from this meeting transcript ·
        ${{ tasks.transcript.output }}
      schema:                          # the contract the model must satisfy
        type: object
        required: [actions]
        properties:
          actions:
            type: array
            items:
              type: object
              required: [owner, task]
              properties:
                owner: { type: string }
                task: { type: string }
                due: { type: string }

  - id: save
    depends_on: [extract]
    invoke:
      tool: "nika:write"
      args:
        path: "./action-items.json"
        content: "${{ tasks.extract.output.actions }}"

  - id: trace
    depends_on: [extract]
    invoke:
      tool: "nika:log"
      args:
        level: info
        message: "Action items extracted to ./action-items.json"

outputs:
  actions:
    value: ${{ tasks.extract.output.actions }}
    type: array
    description: "Typed action items, tracker-ready"

​
How it works

1

A required, typed input

transcript_path is declared with type: string · required: true — the engine validates inputs before any task runs, and the workflow becomes callable with a real contract.
2

The schema IS the contract

infer.schema: is standard JSON Schema. The model must return a matching object — on violation the engine rejects (and may retry). Downstream tasks index into it directly: ${{ tasks.extract.output.actions }}.
3

Two consumers, no extra ceremony

save writes the JSON, trace logs a human-readable line — both depend on extract, both run in parallel.

​
Constructs you just used

ConstructWhereReference
typed vars:transcript_pathYAML syntax
infer.schema:extractThe 4 verbs
schema-path indexingsave.args.contentBindings
nika:logtraceBuiltins

​
Make it yours

  • Add due to the required: list if your team always sets deadlines.
  • Chain a nika:notify task that posts each owner their items.
  • Swap the transcript file for a nika:fetch of your meeting tool’s export API — the schema doesn’t change.

Next · Price watch

The next starter has zero model calls — pure tools, bindings and one CEL comparison.