agentleFS
Sign inStart free
NVIDIAKnown publisherNVIDIA/cdf-curation-and-retrieval/CLAUDE.md

cdf-curation-and-retrieval

CLAUDE.md0 starsChanged 27 days agoApache-2.0View on GitHub

Product / repo: cdf-curation-and-retrieval Integration and operator environment for CDF Curation and Retrieval. Pipeline logic stays in vendor images (Cosmos Curator and Data Mining). This repository exposes Make as the supported local operator UX. The installed cdf_curation_and_retrieval CLI is a non-interactive helper for Make targets, orchestrators, and tests; do not present it as a separate public product UX. Customer documentation: docs/user-guide/README.md. Cookbooks: cookbook/README.md. Default Curator / Data Mining / CDS image pins: root Makefile (override via .env / local.env). Make…

  • Reads credentials

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 1

    1. NVIDIA/paidf-curation-and-retrieval★ 0100% same

    Edited · 0

    None found.

    Wrote their own version · 0

    None found.

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository2 other files it gives its agents.

The file

We split this CLAUDE.md into the rules it gives an agent

7 rules. About 0 are common, 2 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    Run engines via Make:

    line 38
  2. Very specific:

    Default Curator / Data Mining / CDS image pins: root Makefile (override via .env / local.env).

    line 68
  3. Very specific:

    Keep operator documentation Make-first.

    line 84
  4. Very specific:

    Curator: prefer flat YAML with snake_case keys matching upstream argparse dest names.

    line 103
  5. Distinctive:

    Never hardcode machine-specific paths in config YAMLs.

    line 147
  6. Very specific:

    Commit messages follow conventional commits (feat:, fix:, chore:, docs:).

    line 148
  7. Very specific:

    Never commit secrets, local.env, or private registry hosts.

    line 152

≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
CLAUDE.md

Cosmos Data Factory — Curation and Retrieval

Product / repo: cdf-curation-and-retrieval

Integration and operator environment for CDF Curation and Retrieval. Pipeline logic stays in vendor images (Cosmos Curator and Data Mining). This repository exposes Make as the supported local operator UX. The installed cdf_curation_and_retrieval CLI is a non-interactive helper for Make targets, orchestrators, and tests; do not present it as a separate public product UX.

Project Layout

.env.example               Image pins
adapters/                  Curator, Data Mining, and Docker runtime integration
apps/                      Click delivery, workflow modules, composition root
packages/                  Domain types, ports, analytics
configs/                   Full Curator reference YAMLs (split, dedup, shard, image)
cookbook/                  Scenario recipes
docs/                      Operator user-guide and architecture
skills/                    Agent skills (SKILL.md per skill)
tests/unit/                Unit tests
.external/                 Ignored optional upstream source checkouts

Customer documentation: docs/user-guide/README.md.

How It Works

  1. Copy env and sync deps:
    • cp .env.example .env
    • uv sync
  2. Pull vendor images (never build them in the normal path):
    • make pull — Cosmos Curator
    • make pull-data-mining — Data Mining
  3. Curator host FFmpeg sidecar: make ffmpeg-install then make check-setup
  4. Run engines via Make:
    • Curator: make run-pipeline CONFIG_FILE=… / make run_image_pipeline
    • Image embeddings: make run-image-embeddings …
    • Text embeddings: make run-text-embeddings …
    • Mining: make run-data-mining-select TMM_CONFIG_FILE=…
    • Unique match: make run-data-mining-unique-match TMM_UNM_CONFIG_FILE=…

Cookbooks: cookbook/README.md.

Operator Entry Points (Make)

CommandPurpose
make helpShow targets and configuration
make check-setupPrerequisites (Docker, NVIDIA, FFmpeg sidecar)
make pullPull Curator image and retag locally
make pull-data-miningPull tao-toolkit image
make ffmpeg-installInstall host FFmpeg sidecar (conda-forge LGPL)
make download-modelsDownload Curator model weights (~90GB)
make run-pipelineRun Curator video pipeline in Docker
make run_image_pipelineRun Curator image annotate (upstream config-file mode)
make shellInteractive shell in Curator container
make run-data-mining-selectTAO tmm nearest_neighbors
make run-data-mining-unique-matchTAO tmm unique_neighbor_matching
make run-image-embeddingsTAO embedding image_embeddings
make run-text-embeddingsTAO embedding text_embeddings
make formatFormat with ruff
make cleanClean up (preserves images and models)
make clone-curator / make buildOptional Curator source build only

Default Curator / Data Mining / CDS image pins: root Makefile (override via .env / local.env).

Internal Automation CLI

Make targets call the installed helper:

uv run cdf_curation_and_retrieval ...

Canonical helper groups are:

  • cdf_curation_and_retrieval integration ... for finite TAO Data Services image-embedding handoffs.

Keep operator documentation Make-first. Use direct cdf_curation_and_retrieval commands only for automation, tests, or orchestration contracts.

Development

uv sync --extra dev
uv run ruff check .
uv run ruff format .
uv run pytest tests/unit -q
  • Python 3.10+, ruff line-length 100, double quotes.
  • Engines ship inside pulled Docker images; do not redeclare engine-only packages in this project's dependencies.

Configuration

Curator: prefer flat YAML with snake_case keys matching upstream argparse dest names. CDF preflight also accepts a nested args mapping and flattens it before invoking Curator. Upstream fills omitted keys from parser defaults.

Templates in configs/: split.yaml, dedup.yaml, shard.yaml, image.yaml.

Cookbook scenarios under cookbook/ hold minimal recipe overrides.

Mining: Make targets + TAO experiment YAML (tmm.yaml) — see cookbook/.

Environment Variables

Image pins: copy .env.example → .env.

Object-store credentials are injected at runtime by the selected operator deployment mechanism. Handoff commands accept credential environment-variable names (OBJECT_STORE_ACCESS_KEY_ENV, OBJECT_STORE_SECRET_KEY_ENV) rather than credential values. Never commit or document secret values.

Model weights: set model_weights_path in Curator config or MODELS_DIR.

Skills

Skills are in skills/. Each has a SKILL.md with progressive disclosure via references/ subdirectories.

SkillPurpose
cdf-curation-and-retrievalCosmos Curator split / dedup / shard / image annotate: configure, run, troubleshoot, KPI and distribution-aware curation

Scope notes:

  • Data Mining ops are covered by make help / cookbooks, not by expanding this Curator-focused skill into a second product surface.

Domain Catalogs

Domain catalogs in .agents/references/catalogs/ define canonical event taxonomies with tiered severity levels, alias remap tables, and keyword lists. Supported domains: traffic safety, warehouse, construction, parking facility video analytics, retail store video analytics, facility incident analytics, logistics, and workplace visual safety.

Conventions

  • Never hardcode machine-specific paths in config YAMLs; use placeholders.
  • Commit messages follow conventional commits (feat:, fix:, chore:, docs:).
  • Operator UX is Make-first; keep cdf_curation_and_retrieval documented as an internal automation helper, not a separate public UX.
  • Never commit secrets, local.env, or private registry hosts.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free