Concrete Runs

Examples grounded in the actual CLI, workflow templates, and package shape.

These are not invented marketing scenarios. They come from the repo’s real commands, workflow defaults, and output packaging.

Workflow 1

Research a potentially viral Android app opportunity

android-opportunity

Use this when you want the system to scan competitor pages, product communities, app reviews, and user pain points to produce an evidence-backed concept shortlist.

npm run start -- workflow run android-opportunity \
  --topic "ai study planner" \
  --preset deep

What the run should give you

Repeated user pains and feature gaps
Competitor blind spots and monetization clues
A shortlist of app concepts and launch hooks
A report package you can review locally

Why the template helps

The template sets larger research budgets than a casual run and shapes the instruction around app opportunity discovery instead of a generic summary task.

Workflow 2

Build a technical article brief from current web discussion

article-research

Use this when you want to write about a technical topic that is actively discussed on the web and you need both consensus and disagreement captured before drafting.

npm run start -- workflow run article-research \
  --topic "browser automation with Lightpanda and CDP"

What the run should give you

Evidence-backed key findings
Contradictions or disagreements to handle carefully
Article angles and a practical outline
A claim checklist before drafting the final piece

Why the template helps

The workflow brief and handoff files make it much easier to move from research into article production without rereading every raw source.

Queued Run

Queue a long research job and process it with a local worker

queue + worker

Use this when you want the job to sit in durable queue storage first, then let a worker claim it, heartbeat it, and recover it later if the run is interrupted.

npm run start -- workflow enqueue android-opportunity \
  --topic "expense tracking for local merchants"
npm run start -- worker run --once
npm run start -- queue list

Operational advantages

Delayed execution and durable queue rows
Pause, resume, cancel, retry, and rerun controls
Lease recovery for stale or interrupted executions
Job and queue inspection through the dashboard or CLI

Where to inspect it

npm run start -- server run --port 4317

Then open the local dashboard to inspect queue state, recoverable runs, and live logs.

General Agent Run

Run a one-off research instruction outside the shipped templates

agent run

Use this when the task does not fit the built-in templates but you still want the staged pipeline, persisted evidence, and report generation.

npm run start -- agent run \
  "Research cheerful launch ideas for our product and write one evidence-backed post"

Best for

Custom product or market questions
Exploratory research before deciding on a workflow
Shorter evidence-backed writeups

Still included

Planning and research queries
Persisted sources, documents, and evidence
Summary, drafts, and report artifacts

Reference Package

The site also ships a copied workflow package example.

Package README Example

# Technical Article Research Package

Workflow: Technical Article Research
Topic: AI summary automation
Preset: standard
Status: waiting_review

## Operator Flow

- Read the workflow brief first for the short decision-ready version.
- Open the final report when you need evidence detail and source references.
- Use the raw research snapshots only when you need to audit or extend the run.

Manifest Example

{
  "workflowId": "article-research",
  "topic": "AI summary automation",
  "preset": "standard",
  "status": "waiting_review",
  "files": {
    "researchSummary": "handoff/research-summary.md",
    "workflowBrief": "handoff/workflow-brief.md",
    "runtimeManifest": "runtime/pipeline-manifest.json",
    "promptTraces": "runtime/llm-prompt-traces.json"
  }
}

Open the docs for full setup, storage, queue, API, and roadmap details.

Documentation Queue & Controls