nika: v1
workflow:
  id: deep-research-brief
  description: "plan → budgeted research agent → thinking synthesis → brief on disk"

model: ollama/qwen3.5:4b   # local default · per-task overrides below pick stronger models

vars:
  topic:
    type: string
    required: true
    description: "What to research"

tasks:
  plan:
    infer:
      model: anthropic/claude-haiku-4-5    # planning is a fast-model job
      prompt: "Break '${{ vars.topic }}' into 4 sharp research queries · no overlap."
      schema:
        type: object
        required: [queries]
        properties:
          queries:
            type: array
            items: { type: string }

  investigate:
    with:
      plan_queries: ${{ tasks.plan.output.queries }}
    agent:
      system: |
        You are a rigorous researcher. Work the queries one by one ·
        fetch sources · keep verbatim quotes · note what each source
        actually supports. Call nika:done when the plan is exhausted.
      prompt: "Research plan · ${{ with.plan_queries }}"
      tools:
        - "nika:fetch"                 # the web
        - "nika:write"                 # scratch notes
        - "nika:done"                  # the loop-completion sentinel
      max_turns: 25
      max_tokens_total: 150000
      schema:
        type: object
        required: [findings, sources]
        properties:
          findings:
            type: array
            items: { type: string }
          sources:
            type: array
            items: { type: string }

  brief:
    with:
      investigate_findings: ${{ tasks.investigate.output.findings }}
      investigate_sources: ${{ tasks.investigate.output.sources }}
    infer:
      prompt: |
        Findings · ${{ with.investigate_findings }}
        Sources · ${{ with.investigate_sources }}
        Write the executive brief · what's true · what's contested ·
        what it means · what to do. Two pages max.
      thinking:
        enabled: true
        budget_tokens: 8000

  save:
    with:
      brief: ${{ tasks.brief.output }}
    invoke:
      tool: "nika:write"
      args:
        path: "./research/${{ vars.topic }}.md"
        content: "${{ with.brief }}"
        create_dirs: true

outputs:
  brief: ${{ tasks.brief.output }}
  sources:
    value: ${{ tasks.investigate.output.sources }}
    description: "Every source the agent actually used"

{"a":{"graph_format":2,"workflow":"deep-research-brief","nodes":[{"id":"plan","verb":"infer","model":"anthropic/claude-haiku-4-5","permits":[]},{"id":"investigate","verb":"agent","model":"ollama/qwen3.5:4b","permits":["tool: nika:done","tool: nika:fetch","tool: nika:write"]},{"id":"brief","verb":"infer","model":"ollama/qwen3.5:4b","permits":[]},{"id":"save","verb":"invoke","tool":"nika:write","permits":["tool: nika:write"]}],"edges":[{"from":"brief","to":"save","kind":"value"},{"from":"investigate","to":"brief","kind":"value"},{"from":"investigate","to":"brief","kind":"value"},{"from":"plan","to":"investigate","kind":"value"}]},"engine":"0.105.0"}

showcase · tier 4 · research

Deep research brief

A fast model writes the plan, an agent works the web inside hard budgets, a thinking model writes the brief.

“Get me up to speed by Thursday” becomes a pipeline you can audit end to end. · conformance-gated in nika-spec ↗ · re-proven at every push

The whole file

4 tasks · 4 waves · the plan falls out of the bindings, nothing is scheduled by hand.

t4-deep-research-brief.nika.yamlsource

1nika: v12workflow:3  id: deep-research-brief4  description: "plan → budgeted research agent → thinking synthesis → brief on disk"56model: ollama/qwen3.5:4b   # local default · per-task overrides below pick stronger models78vars:9  topic:10    type: string11    required: true12    description: "What to research"1314tasks:15  plan:16    ◇infer:17      model: anthropic/claude-haiku-4-5    # planning is a fast-model job18      prompt: "Break '${{ vars.topic }}' into 4 sharp research queries · no overlap."19      schema:20        type: object21        required: [queries]22        properties:23          queries:24            type: array25            items: { type: string }2627  investigate:28    with:29      plan_queries: ${{ tasks.plan.output.queries }}30    ✦agent:31      system: |32        You are a rigorous researcher. Work the queries one by one ·33        fetch sources · keep verbatim quotes · note what each source34        actually supports. Call nika:done when the plan is exhausted.35      prompt: "Research plan · ${{ with.plan_queries }}"36      tools:37        - "nika:fetch"                 # the web38        - "nika:write"                 # scratch notes39        - "nika:done"                  # the loop-completion sentinel40      max_turns: 2541      max_tokens_total: 15000042      schema:43        type: object44        required: [findings, sources]45        properties:46          findings:47            type: array48            items: { type: string }49          sources:50            type: array51            items: { type: string }5253  brief:54    with:55      investigate_findings: ${{ tasks.investigate.output.findings }}56      investigate_sources: ${{ tasks.investigate.output.sources }}57    ◇infer:58      prompt: |59        Findings · ${{ with.investigate_findings }}60        Sources · ${{ with.investigate_sources }}61        Write the executive brief · what's true · what's contested ·62        what it means · what to do. Two pages max.63      thinking:64        enabled: true65        budget_tokens: 80006667  save:68    with:69      brief: ${{ tasks.brief.output }}70    ◆invoke:71      tool: "nika:write"72      args:73        path: "./research/${{ vars.topic }}.md"74        content: "${{ with.brief }}"75        create_dirs: true7677outputs:78  brief: ${{ tasks.brief.output }}79  sources:80    value: ${{ tasks.investigate.output.sources }}81    description: "Every source the agent actually used"82

nika inspect · engine 0.105.0 · vendored graph, never re-derived

4 declared edges · 4 waves · the plan falls out of the bindings

try itopen in the playground guidethe walkthrough · docs verbinfer verbagent verbinvoke toolnika:write