PAIDF Auto-Labeling
Use this skill when a user wants to kick off PAIDF Auto-Labeling on their
own data, domain, or use case, or when the request matches a shipped cookbook,
stage, authoring, or migration task. This is a router: sequence the specialized
references instead of duplicating their detail.
Routing (Read First)
| Request looks like |
Read |
| New user, clean checkout, first validated run, "how do I get started" |
This file, then the matching reference below |
| Choose annotation targets / stage subset for a domain |
references/scenario-planning.md |
| Create, review, or adapt a cookbook |
references/cookbook-authoring.md |
| Write or adapt VLM/LLM prompts or question banks |
references/prompt-authoring.md |
| Migrate an existing annotation repo into this one |
references/pipeline-migration.md |
| Run the video data augmentation cookbook |
references/video-data-augmentation.md |
| Run or choose an EPAS / PAS cookbook |
references/event-and-person-attribute-search.md |
| Run event-verification reasoning |
references/event-verification-reasoning.md |
| Debug an already-integrated workflow |
references/workflow-runner-debugging.md |
| Implement or review a new stage or Dockerized service |
references/workflow-stage-integration.md |
| Configure or debug one production stage |
The matching file under references/stages/ |
Stage references: super-resolution,
detection-and-tracking,
captioning,
visual-qa,
reasoning,
person-attribute-search,
grounding-2d,
referring-expressions,
training-export.
Instructions
- Confirm the critical run inputs with the user before doing anything else, and
ask a concise question whenever one is missing or ambiguous - never guess or
silently invent a default. At minimum confirm: input data path, output path,
VLM/LLM endpoint URLs and model names, model cache path, GPU ids, and (for
reasoning-capable models) the
max_tokens cap. Restate the confirmed values
back to the user before the first execution.
- Verify the environment: repository cloned,
make targets available, the
model cache path exists, the VLM/LLM endpoints are reachable, and a GPU is
available. State any missing prerequisite as a blocker instead of assuming it.
- Run a shipped example first to confirm the stack works end to end before
customizing. Pick the closest operator pipeline - video data augmentation,
event-and-person-attribute-search, or event-verification-reasoning - and
run its committed cookbook. Use the matching operator reference.
- Plan the target scenario: define modality, domain, intended consumer, and
required annotations, and get a minimal stage subset. Use
scenario-planning.
- Adapt the closest shipped cookbook to the new domain rather than authoring
from scratch. Use cookbook-authoring.
- Author the domain prompts and question banks. Use
prompt-authoring.
- Configure the per-stage settings for the domain (detector classes or SAM3
prompts, endpoints, windowing,
max_tokens). Use the relevant stage
reference, starting with
detection-and-tracking.
- Dry-run the adapted cookbook, then execute and validate the outputs. Use
workflow-runner-debugging.
Adopting an existing external annotation or dataset-generation repository into
PAIDF instead of starting from a shipped cookbook is a migration task; use
pipeline-migration for that path.
Examples
New user, new domain: "I cloned the repo and have my own warehouse-safety video.
How do I produce auto-labels for my domain?"
Guided path:
- Confirm env (model cache, VLM/LLM endpoints, GPU), then prove the stack on a
shipped example before customizing:
make run SCRIPT=workflow-runner:main \
ARGS='--cookbook-file cookbooks/video_data_augmentation/configs/pipeline_video.yaml --container-dry-run'
- Plan the domain (scenario-planning) -> subset
detection_and_tracking -> captioning -> visual_qa -> reasoning -> training_export
(add grounding_2d for caption→boxes or referring_expressions for boxes→phrases;
use grounding-2d /
referring-expressions).
- Copy the closest cookbook to
cookbooks/warehouse_safety/configs/pipeline.yaml
and adapt inputs, detector classes/SAM3 prompts, prompts, and question banks.
- Dry-run the new cookbook, then run for real and validate outputs:
make run SCRIPT=workflow-runner:main \
ARGS='--cookbook-file cookbooks/warehouse_safety/configs/pipeline.yaml --container-dry-run'
Guardrails
- Do not guess or fabricate the critical inputs enumerated in step 1; if any is
missing or ambiguous, ask the user and confirm before executing.
- Do not customize a cookbook before a shipped example runs clean; a broken base
makes domain debugging ambiguous.
- Keep the first custom pipeline minimal - only the stages needed for the
requested annotations - and expand later.
- Verify that every selected stage's service package and image exist in the
current branch before promising an end-to-end run.
- Do not put secrets, tokens, or absolute home paths in committed cookbooks; use
placeholders such as
<model-cache> and env vars for endpoint keys.
- For reasoning-capable models (for example Gemini 3 Flash), raise
max_tokens
on the visual_qa and reasoning LLM substages to avoid the thinking-token
tax; keep the default cap for non-reasoning models.
- Do not rely on non-PAIDF pipelines, commands, or file locations. A first run
must be reproducible through
workflow-runner:main inside this repo.
1---2name: paidf-auto-labeling3description: >-4 Use when a user needs to get started with PAIDF Auto-Labeling, plan a5 scenario, run or debug a shipped cookbook, author prompts or cookbooks,6 migrate a pipeline, or configure a stage. Confirm critical inputs (data7 path, output path, endpoints) and ask when any are missing. This is a8 router: read the matching reference instead of inventing a workflow.9license: Apache-2.010owner: NVIDIA11service: physical-ai-data-factory12version: 1.1.013reviewed: 2026-08-2514author: "NVIDIA <opensource@nvidia.com>"15metadata:16 source: https://skillmd.com/skills/nvidia/paidf-auto-labeling17 author: "NVIDIA <opensource@nvidia.com>"18 tags: [getting-started, onboarding, new-domain, quickstart, new-use-case]19---2021# PAIDF Auto-Labeling2223Use this skill when a user wants to kick off PAIDF Auto-Labeling on their24own data, domain, or use case, or when the request matches a shipped cookbook,25stage, authoring, or migration task. This is a router: sequence the specialized26references instead of duplicating their detail.2728## Routing (Read First)2930| Request looks like | Read |31| --- | --- |32| New user, clean checkout, first validated run, "how do I get started" | This file, then the matching reference below |33| Choose annotation targets / stage subset for a domain | [`references/scenario-planning.md`](references/scenario-planning.md) |34| Create, review, or adapt a cookbook | [`references/cookbook-authoring.md`](references/cookbook-authoring.md) |35| Write or adapt VLM/LLM prompts or question banks | [`references/prompt-authoring.md`](references/prompt-authoring.md) |36| Migrate an existing annotation repo into this one | [`references/pipeline-migration.md`](references/pipeline-migration.md) |37| Run the video data augmentation cookbook | [`references/video-data-augmentation.md`](references/video-data-augmentation.md) |38| Run or choose an EPAS / PAS cookbook | [`references/event-and-person-attribute-search.md`](references/event-and-person-attribute-search.md) |39| Run event-verification reasoning | [`references/event-verification-reasoning.md`](references/event-verification-reasoning.md) |40| Debug an already-integrated workflow | [`references/workflow-runner-debugging.md`](references/workflow-runner-debugging.md) |41| Implement or review a new stage or Dockerized service | [`references/workflow-stage-integration.md`](references/workflow-stage-integration.md) |42| Configure or debug one production stage | The matching file under [`references/stages/`](references/stages/) |4344Stage references: [super-resolution](references/stages/super-resolution.md),45[detection-and-tracking](references/stages/detection-and-tracking.md),46[captioning](references/stages/captioning.md),47[visual-qa](references/stages/visual-qa.md),48[reasoning](references/stages/reasoning.md),49[person-attribute-search](references/stages/person-attribute-search.md),50[grounding-2d](references/stages/grounding-2d.md),51[referring-expressions](references/stages/referring-expressions.md),52[training-export](references/stages/training-export.md).5354## Instructions55561. Confirm the critical run inputs with the user before doing anything else, and57 ask a concise question whenever one is missing or ambiguous - never guess or58 silently invent a default. At minimum confirm: input data path, output path,59 VLM/LLM endpoint URLs and model names, model cache path, GPU ids, and (for60 reasoning-capable models) the `max_tokens` cap. Restate the confirmed values61 back to the user before the first execution.622. Verify the environment: repository cloned, `make` targets available, the63 model cache path exists, the VLM/LLM endpoints are reachable, and a GPU is64 available. State any missing prerequisite as a blocker instead of assuming it.653. Run a shipped example first to confirm the stack works end to end before66 customizing. Pick the closest operator pipeline - video data augmentation,67 event-and-person-attribute-search, or event-verification-reasoning - and68 run its committed cookbook. Use the matching operator reference.694. Plan the target scenario: define modality, domain, intended consumer, and70 required annotations, and get a minimal stage subset. Use71 [scenario-planning](references/scenario-planning.md).725. Adapt the closest shipped cookbook to the new domain rather than authoring73 from scratch. Use [cookbook-authoring](references/cookbook-authoring.md).746. Author the domain prompts and question banks. Use75 [prompt-authoring](references/prompt-authoring.md).767. Configure the per-stage settings for the domain (detector classes or SAM377 prompts, endpoints, windowing, `max_tokens`). Use the relevant stage78 reference, starting with79 [detection-and-tracking](references/stages/detection-and-tracking.md).808. Dry-run the adapted cookbook, then execute and validate the outputs. Use81 [workflow-runner-debugging](references/workflow-runner-debugging.md).8283Adopting an existing external annotation or dataset-generation repository into84PAIDF instead of starting from a shipped cookbook is a migration task; use85[pipeline-migration](references/pipeline-migration.md) for that path.8687## Examples8889New user, new domain: "I cloned the repo and have my own warehouse-safety video.90How do I produce auto-labels for my domain?"9192Guided path:9394- Confirm env (model cache, VLM/LLM endpoints, GPU), then prove the stack on a95 shipped example before customizing:9697```bash98make run SCRIPT=workflow-runner:main \99 ARGS='--cookbook-file cookbooks/video_data_augmentation/configs/pipeline_video.yaml --container-dry-run'100```101102- Plan the domain ([scenario-planning](references/scenario-planning.md)) -> subset103 `detection_and_tracking -> captioning -> visual_qa -> reasoning -> training_export`104 (add `grounding_2d` for caption→boxes or `referring_expressions` for boxes→phrases;105 use [grounding-2d](references/stages/grounding-2d.md) /106 [referring-expressions](references/stages/referring-expressions.md)).107- Copy the closest cookbook to `cookbooks/warehouse_safety/configs/pipeline.yaml`108 and adapt inputs, detector classes/SAM3 prompts, prompts, and question banks.109- Dry-run the new cookbook, then run for real and validate outputs:110111```bash112make run SCRIPT=workflow-runner:main \113 ARGS='--cookbook-file cookbooks/warehouse_safety/configs/pipeline.yaml --container-dry-run'114```115116## Guardrails117118- Do not guess or fabricate the critical inputs enumerated in step 1; if any is119 missing or ambiguous, ask the user and confirm before executing.120- Do not customize a cookbook before a shipped example runs clean; a broken base121 makes domain debugging ambiguous.122- Keep the first custom pipeline minimal - only the stages needed for the123 requested annotations - and expand later.124- Verify that every selected stage's service package and image exist in the125 current branch before promising an end-to-end run.126- Do not put secrets, tokens, or absolute home paths in committed cookbooks; use127 placeholders such as `<model-cache>` and env vars for endpoint keys.128- For reasoning-capable models (for example Gemini 3 Flash), raise `max_tokens`129 on the `visual_qa` and `reasoning` LLM substages to avoid the thinking-token130 tax; keep the default cap for non-reasoning models.131- Do not rely on non-PAIDF pipelines, commands, or file locations. A first run132 must be reproducible through `workflow-runner:main` inside this repo.