This guide helps you choose a checked-in OtterDesk blueprint for a first run, smoke test, or runtime demonstration.
Run commands from the workspace root unless a step says otherwise.
mn blueprint listExpected output includes:
ID
Name
Job Name
Catalog availability depends on your cached blueprint repository. The checked-in
local catalog for this workspace is otterdesk-blueprints/index.json.
| Order | Blueprint | Use it when |
|---|---|---|
| 1 | tax_form_ocr_capture_assistant |
You want a small document/OCR workflow with public sample metadata. |
| 2 | portfolio_risk_review_assistant |
You want a finance workflow with deterministic simulation and LLM report writing. |
| 3 | video_watch_assistant |
You want a service-style video monitoring workflow. |
| 4 | gtm_ai_workflow |
You want a long-running GTM automation loop with CRM and outreach artifacts. |
| 5 | personal_financial_advisor |
You want a folder-watching service with OCR, public browser research, and review-only reports. |
| 6 | generic_customer_service_voice_coworker |
You want a local voice service backed by the NVIDIA/Spark stack. |
Path:
otterdesk-blueprints/tax_form_ocr_capture_assistant
Use it when:
- you want a compact checked-in blueprint;
- you want document intake and OCR-style artifact output;
- you want a review-only workflow with public dataset notes.
Run:
mn blueprint run --folder otterdesk-blueprints/tax_form_ocr_capture_assistantDirect runner smoke test:
cd otterdesk-blueprints/tax_form_ocr_capture_assistant
python3.11 payloads/document_workflow/scripts/run_blueprint.py --runs-root /tmp/mn-runs --run-id tax-form-demoPath:
otterdesk-blueprints/portfolio_risk_review_assistant
Use it when:
- you want public market data plus deterministic risk simulation;
- you want the LLM limited to interpretation and report writing;
- you want a clear review-only human approval policy.
Run from the catalog:
mn blueprint run portfolio_risk_review_assistantRun from the local folder:
mn blueprint run --folder otterdesk-blueprints/portfolio_risk_review_assistantPath:
otterdesk-blueprints/video_watch_assistant
Use it when:
- you want a service blueprint;
- you want visual detection events and cooldown state;
- you want to inspect live run status and service artifacts.
Run:
mn blueprint run --folder otterdesk-blueprints/video_watch_assistant --web-uiCancel the service when done:
mn job cancel <job_id>Path:
otterdesk-blueprints/gtm_ai_workflow
Use it when:
- you want a long-running account research and outreach loop;
- you want local CSV CRM and market-insight artifacts;
- you want to test email/inbox integrations in dry-run or test-recipient mode first.
Run:
mn blueprint run --folder otterdesk-blueprints/gtm_ai_workflowInspect run state:
mn blueprint monitor --followPath:
otterdesk-blueprints/personal_financial_advisor
Use it when:
- you want a continuous local folder-watch service;
- you want OCR plus public, privacy-safe browser research;
- you want review-only household finance reports and risk reminders.
Bounded direct runner smoke test:
cd otterdesk-blueprints/personal_financial_advisor
python3.11 payloads/document_workflow/scripts/run_blueprint.py --runs-root /tmp/mn-runs --run-id personal-finance-demo --watch --max-cycles 1Path:
otterdesk-blueprints/generic_customer_service_voice_coworker
Use it when:
- you want a real-time voice service;
- you have the local NVIDIA/Spark stack ready;
- you want editable menu knowledge and a localhost HTTPS voice page.
Run with longer cold-start timeouts when the NVIDIA stack needs to warm up:
MN_PRE_LAUNCH_TIMEOUT_SECONDS=900 NEMOTRON_PRELAUNCH_WAIT_SECONDS=900 mn blueprint run generic_customer_service_voice_coworker- Review
manifest.jsonbefore running any blueprint. - Check
pass_envbefore passing secrets. - Use mock, sample, quick-test, and dry-run modes before external delivery.
- Treat financial, tax, legal, healthcare, and safety outputs as review-only until a qualified human approves them.
- Cancel service jobs when you are done.