{"id":50892980,"url":"https://github.com/spiraldb/spiraldb-nemo-curator","last_synced_at":"2026-06-15T22:01:56.092Z","repository":{"id":359630566,"uuid":"1244860576","full_name":"spiraldb/spiraldb-nemo-curator","owner":"spiraldb","description":"SpiralDB connectors for NVIDIA NeMo Curator","archived":false,"fork":false,"pushed_at":"2026-05-22T18:00:51.000Z","size":274,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"develop","last_synced_at":"2026-05-22T21:31:49.461Z","etag":null,"topics":["computer-vision","data","data-curation","data-prep","data-preparation","data-processing","data-quality","datacuration","datarecipes","deduplication","fast-data-processing","multimodal","multimodal-ai","nvidia-nemo","physical-ai","python","spiral","vortex"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":false,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/spiraldb.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-05-20T17:08:49.000Z","updated_at":"2026-05-22T18:00:55.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/spiraldb/spiraldb-nemo-curator","commit_stats":null,"previous_names":["spiraldb/spiraldb-nemo-curator"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/spiraldb/spiraldb-nemo-curator","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/spiraldb%2Fspiraldb-nemo-curator","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/spiraldb%2Fspiraldb-nemo-curator/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/spiraldb%2Fspiraldb-nemo-curator/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/spiraldb%2Fspiraldb-nemo-curator/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/spiraldb","download_url":"https://codeload.github.com/spiraldb/spiraldb-nemo-curator/tar.gz/refs/heads/develop","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/spiraldb%2Fspiraldb-nemo-curator/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34381762,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-15T02:00:07.085Z","response_time":63,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["computer-vision","data","data-curation","data-prep","data-preparation","data-processing","data-quality","datacuration","datarecipes","deduplication","fast-data-processing","multimodal","multimodal-ai","nvidia-nemo","physical-ai","python","spiral","vortex"],"created_at":"2026-06-15T22:01:54.030Z","updated_at":"2026-06-15T22:01:56.073Z","avatar_url":"https://github.com/spiraldb.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"[![CI](https://github.com/spiraldb/spiraldb-nemo-curator/actions/workflows/ci.yml/badge.svg?branch=develop)](https://github.com/spiraldb/spiraldb-nemo-curator/actions/workflows/ci.yml?query=branch%3Adevelop)\n[![License: Apache 2.0](https://img.shields.io/badge/license-Apache--2.0-blue.svg)](LICENSE)\n\n# SpiralDB source for NVIDIA NeMo Curator\n\n\nSpiralDB-backed I/O endpoints for [NVIDIA NeMo Curator][nemo] video pipelines.\nThis library provides two drop-in replacements for Curator's file-based I/O\nstages:\n\n| this library         | replaces                                                                                                                                            | role                                            |\n| -------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------- |\n| `SpiralVideoReader`  | [`VideoReader`](https://docs.nvidia.com/nemo/curator/nemo-curator/nemo_curator/stages/video/io/video_reader#nemo_curator-stages-video-io-video_reader-VideoReader)        | reads source MP4s into `VideoTask`s             |\n| `SpiralClipWriter`   | [`ClipWriterStage`](https://docs.nvidia.com/nemo/curator/latest/nemo-curator/nemo_curator/stages/video/io/clip_writer#nemo_curator-stages-video-io-clip_writer-ClipWriterStage) | writes per-clip outputs at the end of the pipeline |\n\nSwap those two stages in and your raw MP4s, per-clip MP4s, previews, captions,\nand embeddings all live in a single SpiralDB table — no S3 prefixes to babysit,\nno JSON sidecar files, no Parquet shards to compact.\n\nThe middle of the pipeline (clip extraction, motion / aesthetic filtering,\nCosmos-Embed1, captioning) is reused unchanged from `nemo_curator`. Only the\nendpoints change.\n\n## Install\n\n```bash\nuv init my-curation-project\ncd my-curation-project\nuv add nemo-curator git+https://github.com/spiraldb/spiraldb-nemo-curator\n```\n\n## Quickstart\n\nAssumes a Spiral project with an `input_videos` table whose key column is\n`clip_id` and whose `video` column is an `se.Blob` of MP4 bytes. The `clips`\noutput table is auto-created on first run.\n\n```python\nimport os\n\nfrom nemo_curator.pipeline.pipeline import Pipeline\nfrom nemo_curator.stages.video.clipping.clip_extraction_stages import (\n    FixedStrideExtractorStage,\n)\nfrom nemo_curator.stages.video.clipping.clip_transcoding_stages import (\n    ClipTranscodingStage,\n)\nfrom nemo_curator.stages.video.clipping.frame_extraction import FrameExtractionStage\n\nfrom spiraldb_nemo_curator.io.video import SpiralClipWriter, SpiralVideoReader\n\nproject = os.environ[\"SPIRAL_PROJECT\"]\n\npipeline = Pipeline(name=\"quickstart\")\npipeline.add_stage(\n    SpiralVideoReader(\n        project_id=project,\n        table_name=\"input_videos\",\n        video_column=\"video\",\n        source_id_column=\"clip_id\",\n    )\n)\npipeline.add_stage(FrameExtractionStage())\npipeline.add_stage(FixedStrideExtractorStage())\npipeline.add_stage(ClipTranscodingStage(encoder=\"libopenh264\"))\npipeline.add_stage(\n    SpiralClipWriter(project_id=project, table_name=\"clips\")\n)\n\npipeline.run()\n```\n\nA fuller example with TransNetV2, motion filtering, Cosmos-Embed1, and\ncaptioning lives in [`examples/spiral_pipeline.py`](examples/spiral_pipeline.py).\n\n## Architecture\n\n```mermaid\nflowchart LR\n    Src[(SpiralDB\u003cbr/\u003einput_videos)]\n\n    subgraph reader[\"SpiralVideoReader (this library)\"]\n        direction TB\n        P[SpiralPartitionStage\u003cbr/\u003ekeys-only scan]\n        R[SpiralVideoReaderStage\u003cbr/\u003ekeyed point-scan]\n        P --\u003e R\n    end\n\n    subgraph curator[\"NeMo Curator stages (unchanged)\"]\n        direction TB\n        F[FrameExtraction]\n        C[ClipExtraction\u003cbr/\u003eTransNetV2 / FixedStride]\n        T[ClipTranscoding]\n        M[Motion / Aesthetic filters]\n        E[CosmosEmbed1]\n        Cap[CaptionGeneration]\n        F --\u003e C --\u003e T --\u003e M --\u003e E --\u003e Cap\n    end\n\n    W[SpiralClipWriter\u003cbr/\u003ethis library]\n    Dst[(SpiralDB\u003cbr/\u003eclips)]\n\n    Src --\u003e P\n    R --\u003e F\n    Cap --\u003e W\n    W --\u003e Dst\n```\n\n`SpiralVideoReader` is a [`CompositeStage`][composite-stage] that decomposes\ninto two [`ProcessingStage`s][processing-stage]: a cheap keys-only scan (one\ntask per source row) followed by a keyed point-scan that materializes the blob\npayload — so workers fan out over rows without serializing the source table\nacross the cluster. `SpiralClipWriter` is a single `ProcessingStage` that\nwrites both `video.clips` and `video.filtered_clips` in one batched\n`tbl.write()`, with each `se.Blob` column landing in its own column group.\n\n[composite-stage]: https://docs.nvidia.com/nemo/curator/latest/nemo-curator/nemo_curator/stages/base#nemo_curator-stages-base-CompositeStage\n[processing-stage]: https://docs.nvidia.com/nemo/curator/latest/nemo-curator/nemo_curator/stages/base#nemo_curator-stages-base-ProcessingStage\n\n## Contact\n\nQuestions, design feedback, or want help wiring this into your pipeline?\nEmail [nemo-curator@spiraldb.com](mailto:nemo-curator@spiraldb.com?subject=NeMo%20Curator)\nand mention **NeMo Curator** in the subject line.\n\n[nemo]: https://github.com/NVIDIA/NeMo-Curator\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fspiraldb%2Fspiraldb-nemo-curator","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fspiraldb%2Fspiraldb-nemo-curator","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fspiraldb%2Fspiraldb-nemo-curator/lists"}