New tasks are created when syncing again after export , When using the s3 cloud storage of Label Studio

[chunxueer]

Describe the bug
New tasks are created when syncing again after export , When using the s3 cloud storage of Label Studio

To Reproduce
Steps to reproduce the behavior:

1. Create a project, import images, and annotate them

2.Configure S3 cloud storage, export first and then import

3.There is an additional task in the task list that cannot be annotated, and there are differences in its attributes

4.The files exported from S3 cloud storage are numbered and do not have a. json file suffix

Expected behavior

1. I hope that the number of tasks after exporting and importing remains unchanged.
2. Annotations and images can also be displayed normally.
3. The file name exported from S3 cloud storage data files is not just a number, but a JSON file with the same name as the image.

Environment (please complete the following information):

-- docker-compose.yml

services:
  sti-label-studio:
    stdin_open: true
    tty: true 
    image: heartexlabs/label-studio:latest
    restart: unless-stopped
    expose:
      - "8000"
    ports:
      - "35001:8000" 
    environment:
      - DJANGO_DB=default
      - POSTGRE_NAME=${POSTGRE_NAME}
      - POSTGRE_USER=${POSTGRE_USER}
      - POSTGRE_PASSWORD=${POSTGRE_PASSWORD}
      - POSTGRE_PORT=${POSTGRE_PORT}
      - POSTGRE_HOST=${POSTGRE_HOST}
      - LABEL_STUDIO_HOST=${LABEL_STUDIO_HOST:-}
      
      - AWS_ACCESS_KEY_ID=${AWS_ACCESS_KEY_ID}
      - AWS_SECRET_ACCESS_KEY=${AWS_SECRET_ACCESS_KEY}
      - AWS_STORAGE_BUCKET_NAME=${AWS_STORAGE_BUCKET_NAME}
      - AWS_S3_ENDPOINT_URL=${AWS_S3_ENDPOINT_URL}
      - AWS_S3_REGION_NAME=${AWS_S3_REGION_NAME}
      - AWS_DEFAULT_ACL=${AWS_DEFAULT_ACL}
      - USE_S3_STORAGE=${USE_S3_STORAGE}
      
      - JSON_LOG=1
    #      - LOG_LEVEL=DEBUG
    volumes:
      - /data/sti/label-studio/data:/label-studio/data:rw
    command: label-studio-uwsgi

[chunxueer]

Thanks a lot

[chunxueer]

http://10.20.16.79:35001/version

{
  "release": "1.21.0",
  "label-studio-os-package": {
    "version": "1.21.0",
    "short_version": "1.21",
    "latest_version_from_pypi": "1.21.0",
    "latest_version_upload_time": "2025-09-30T06:48:47",
    "current_version_is_outdated": false
  },

  "label-studio-os-backend": {
    "message": "feat: BROS-353: Databricks storage integration (#8245) (#8542)",
    "commit": "5c988322ab98dc71c4dbc3a2e66159d588038888",
    "date": "2025/09/26 10:27:26",
    "branch": "",
    "version": "1.21.0+0.g5c98832"
  },

  "label-studio-frontend": {
    "message": "fix: BROS-389: Fix waveforms rendering for multi channels case (#8415)",
    "commit": "3d1fc618c10e0de7e161469fcfbde331cd347528",
    "date": "2025-09-16T22:21:39.000Z",
    "branch": "develop"
  },

  "dm2": {
    "message": "fix: FIT-641: 'Label Tasks As Displayed' is displayed outside the Labe ...",
    "commit": "a602b3c0a5b44f2bf7cc8ed05b4972b35e086bd5",
    "date": "2025-09-10T19:09:21.000Z",
    "branch": "develop"
  },

  "label-studio-converter": {
    "version": "2.0.4"
  },

  "edition": "Community"
}

[heidi-humansignal]

Hello,

Thank you for contacting Label Studio,
By default, when Sync is triggered, Label Studio will tend to create tasks based on the imported data. can you please double check if the S3 Target Storage and Source Storage has the same bucket/prefix ? if so, can you please try the following workflow and let me know if it has solved the issue

1. If you only need a backup to S3 (don’t re-import into the same project)

• Keep Target Storage solely for export/backup.
• Do not add that same bucket/prefix back as a Source Storage in the same project.
• If you must re-import later, import into a new project.

Best Regards!

Comment by Oussama Assili
Workflow Run

[chunxueer]

1. Both the imported storage bucket A and the exported storage bucket B are the same one.
2. My scenario is that I want a program to periodically update storage bucket A by adding new to-be-annotated images. After annotations are made in Label Studio, I hope they will be synced to storage bucket B, and then an additional program will be used to train the model.
3. It is acceptable for storage buckets A and B to use different prefixes. The core issue is that when files from storage bucket A are synced over, the annotations are not displayed; instead, they are imported into data.annotations.
4. The exported files are named with keys and do not have user-friendly file names or extensions, such as ${time}_${id}.json

http://10.20.16.79:35001/api/tasks/573?project=22

{
    "id": 573,
    "predictions": [],
    "annotations": [],
    "drafts": [],
    "annotators": [],
    "inner_id": 2,
    "cancelled_annotations": 0,
    "total_annotations": 0,
    "total_predictions": 0,
    "completed_at": null,
    "annotations_results": "",
    "predictions_results": "",
    "predictions_score": null,
    "file_upload": null,
    "storage_filename": "datasets/1214243d1dcc47959000ecbbe6dc2395/artifacts/annotations/24",
    "annotations_ids": "",
    "predictions_model_versions": "",
    "avg_lead_time": null,
    "draft_exists": false,
    "updated_by": [],
    "data": {
        "id": 24,
        "task": {
            "id": 560,
            "data": {
                "meta": {
                    "file_name": "fish1005_jpg.rf.2d11fcb83b85cc9e091781222d714baa.jpg",
                    "label_path": "/tasks/573/resolve/?fileuri=czM6Ly9zdGktbWxmbG93L2RhdGFzZXRzLzEyMTQyNDNkMWRjYzQ3OTU5MDAwZWNiYmU2ZGMyMzk1L2FydGlmYWN0cy90ZXN0L2xhYmVscy9maXNoMTAwNV9qcGcucmYuMmQxMWZjYjgzYjg1Y2M5ZTA5MTc4MTIyMmQ3MTRiYWEudHh0",
                    "label_content": "0 0.5984375 0.2671875 0.553125 0.53125"
                },
                "image": "/tasks/573/resolve/?fileuri=czM6Ly9zdGktbWxmbG93L2RhdGFzZXRzLzEyMTQyNDNkMWRjYzQ3OTU5MDAwZWNiYmU2ZGMyMzk1L2FydGlmYWN0cy90ZXN0L2ltYWdlcy9maXNoMTAwNV9qcGcucmYuMmQxMWZjYjgzYjg1Y2M5ZTA5MTc4MTIyMmQ3MTRiYWEuanBn",
                "annotations": [
                    {
                        "id": 1,
                        "result": [
                            {
                                "id": "960bb078-ec4f-4d85-ac92-f147b4da2555",
                                "type": "rectanglelabels",
                                "value": {
                                    "x": 32.1875,
                                    "y": 0.15625000000000222,
                                    "width": 55.3125,
                                    "height": 53.125,
                                    "rotation": 0,
                                    "rectanglelabels": [
                                        "fishing"
                                    ]
                                },
                                "origin": "imported",
                                "to_name": "image",
                                "from_name": "rectanglelabels",
                                "image_rotation": 0,
                                "original_width": 640,
                                "original_height": 640
                            }
                        ],
                        "imported": true,
                        "lead_time": 0,
                        "created_at": "2025-11-18T17:05:32.988967Z",
                        "prediction": {},
                        "updated_at": "2025-11-18T17:05:32.988967Z",
                        "last_action": "imported",
                        "bulk_created": true,
                        "completed_by": 1,
                        "ground_truth": true,
                        "result_count": 1,
                        "was_cancelled": false,
                        "last_action_by": 1,
                        "draft_created_at": "2025-11-18T17:05:32.988967Z"
                    }
                ]
            },
            "meta": {},
            "overlap": 1,
            "project": 22,
            "inner_id": 1,
            "created_at": "2025-11-18T09:09:59.484222Z",
            "is_labeled": true,
            "updated_at": "2025-11-18T09:14:24.294111Z",
            "updated_by": 1,
            "file_upload": null,
            "comment_count": 0,
            "comment_authors": [],
            "total_annotations": 1,
            "total_predictions": 0,
            "cancelled_annotations": 0,
            "last_comment_updated_at": null,
            "unresolved_comment_count": 0
        },
        "result": [
            {
                "id": "ORB5zmeyJt",
                "type": "rectanglelabels",
                "value": {
                    "x": 30.321592649310876,
                    "y": 11.638591117917304,
                    "width": 29.709035222052062,
                    "height": 42.725880551301685,
                    "rotation": 0,
                    "rectanglelabels": [
                        "fishing"
                    ]
                },
                "origin": "manual",
                "to_name": "image",
                "from_name": "rectanglelabels",
                "image_rotation": 0,
                "original_width": 640,
                "original_height": 640
            }
        ],
        "project": 22,
        "import_id": null,
        "lead_time": 7.191,
        "created_at": "2025-11-18T09:14:24.218645Z",
        "updated_at": "2025-11-18T09:14:24.218665Z",
        "updated_by": 1,
        "created_ago": "12 minutes",
        "last_action": null,
        "bulk_created": false,
        "completed_by": {
            "id": 1,
            "email": "lichun@cnovit.com",
            "last_name": "",
            "first_name": ""
        },
        "ground_truth": false,
        "was_cancelled": false,
        "last_created_by": null,
        "created_username": " lichun@cnovit.com, 1",
        "draft_created_at": null,
        "parent_annotation": null,
        "parent_prediction": null
    },
    "meta": {},
    "created_at": "2025-11-18T09:26:32.616350Z",
    "updated_at": "2025-11-18T09:26:32.616370Z",
    "is_labeled": false,
    "overlap": 1,
    "comment_count": 0,
    "unresolved_comment_count": 0,
    "last_comment_updated_at": null,
    "project": 22,
    "comment_authors": []
}

[chunxueer]

input file

{"id": 27, "result": [{"id": "d9wj-GkZo_", "type": "rectanglelabels", "value": {"x": 28.62669245647969, "y": 7.35009671179884, "width": 27.75628626692456, "height": 45.64796905222436, "rotation": 0, "rectanglelabels": ["fishing"]}, "origin": "manual", "to_name": "image", "from_name": "rectanglelabels", "image_rotation": 0, "original_width": 640, "original_height": 640}], "created_username": " lichun@cnovit.com, 1", "created_ago": "1\u00a0minute", "completed_by": {"id": 1, "first_name": "", "last_name": "", "email": "lichun@cnovit.com"}, "task": {"id": 594, "data": {"meta": {"run_id": "1214243d1dcc47959000ecbbe6dc2395", "file_name": "fish1005_jpg.rf.2d11fcb83b85cc9e091781222d714baa.jpg", "label_path": "s3://sti-mlflow/datasets/1214243d1dcc47959000ecbbe6dc2395/artifacts/test/labels/fish1005_jpg.rf.2d11fcb83b85cc9e091781222d714baa.txt", "upload_order": 2, "label_content": "0 0.5984375 0.2671875 0.553125 0.53125"}, "image": "s3://sti-mlflow/datasets/1214243d1dcc47959000ecbbe6dc2395/artifacts/test/images/fish1005_jpg.rf.2d11fcb83b85cc9e091781222d714baa.jpg", "annotations": [{"id": 1, "result": [{"id": "cf2d14d5-a780-4114-bb1a-014589868d98", "type": "rectanglelabels", "value": {"x": 32.1875, "y": 0.15625000000000222, "width": 55.3125, "height": 53.125, "rotation": 0, "rectanglelabels": ["fishing"]}, "origin": "imported", "to_name": "image", "from_name": "rectanglelabels", "image_rotation": 0, "original_width": 640, "original_height": 640}], "imported": true, "lead_time": 0, "created_at": "2025-11-19T18:42:05.598571Z", "prediction": {}, "updated_at": "2025-11-19T18:42:05.598571Z", "last_action": "imported", "bulk_created": true, "completed_by": 1, "ground_truth": true, "result_count": 1, "was_cancelled": false, "last_action_by": 1, "draft_created_at": "2025-11-19T18:42:05.598571Z"}]}, "meta": {}, "created_at": "2025-11-19T10:42:06.029473Z", "updated_at": "2025-11-19T10:43:13.799554Z", "is_labeled": true, "overlap": 1, "inner_id": 1, "total_annotations": 1, "cancelled_annotations": 0, "total_predictions": 0, "comment_count": 0, "unresolved_comment_count": 0, "last_comment_updated_at": null, "project": 28, "updated_by": 1, "file_upload": null, "comment_authors": []}, "was_cancelled": false, "ground_truth": false, "created_at": "2025-11-19T10:43:13.665624Z", "updated_at": "2025-11-19T10:43:13.665640Z", "draft_created_at": null, "lead_time": 5.183, "import_id": null, "last_action": null, "bulk_created": false, "project": 28, "updated_by": 1, "parent_prediction": null, "parent_annotation": null, "last_created_by": null}

[heidi-humansignal]

Hello,

Thanks for the additional details, Below is a robust setup for your scenario: new images arrive in S3 A, you annotate in Label Studio, and exports land in S3 B for your training job, without duplicates and with readable .json files.

• Target (export) S3 writes objects that are not meant to be re‑imported by default; importing those back creates extra placeholder tasks.
• Your Source storage likely scans both images and exported files, so re‑sync creates new tasks again.

1. Split your S3 prefixes and what each storage is allowed to scan

• Source (A): keep only media under a dedicated prefix, e.g. s3://bucket-a/images/.
• File filter regex: something like .*.(jpg|jpeg|png)$
• Treat every bucket object as a source file: ON
• This creates tasks only from images. Re‑sync is idempotent because LS tracks the S3 key and skips already‑linked files.
• Target (B): write only task JSON to a different prefix, e.g. s3://bucket-b/exports/tasks/.
• Do not point any Source storage at this prefix in the same project.
• This avoids importing your own exports and prevents duplicates.

2. Export per‑task JSON with a .json extension
   By default, Target Storage writes objects named by internal numeric IDs with no extension. To emit per‑task JSON that is re‑importable and has .json:

• Add this env var and restart Label Studio:FUTURE_SAVE_TASK_TO_STORAGE=1
• Click Sync on your Target S3 storage (Project → Settings → Cloud Storage → Target). You’ll get files like <task_id>.json containing the full task (data + annotations).

Best Regards!

Comment by Oussama Assili
Workflow Run