Local File Sink

Class: LocalFileSinkBlockV1

Source: inference.core.workflows.core_steps.sinks.local_file.v1.LocalFileSinkBlockV1

The Local File Sink block saves workflow data as files on a local file system. It allows users to configure how the data is stored, either:

• aggregating multiple entries into a single file

• or saving each entry as a separate file.

This block is useful for logging, data export, or preparing files for subsequent processing.

File Content, File Type and Output Mode

content is expected to be the output from another block producing string values of specific types denoted by file_type.

output_mode set into append_log will make the block appending single file with consecutive entries passed to content input up to max_entries_per_file. In this mode it is important that

file_type in append_log mode

Contrary to separate_files output mode, append_log mode may introduce subtle changes into the structure of the content to properly append it into existing file, hence setting proper file_type is crucial:

• file_type=json: in append_log mode, the block will create *.jsonl file in JSON Lines format - for that to be possible, each JSON document will be parsed and dumped to ensure that it will fit into single line.

• file_type=csv: in append_log mode, the block will deduct the first line from the content (making it required for CSV content to always be shipped with header row) of consecutive updates into the content of already created file.

Security considerations

The block has an ability to write to the file system. If you find this unintended in your system, you can disable the block setting environmental variable ALLOW_WORKFLOW_BLOCKS_ACCESSING_LOCAL_STORAGE=False in the environment which host Workflows Execution Engine.

If you want to restrict the directory which may be used to write data - set environmental variable WORKFLOW_BLOCKS_WRITE_DIRECTORY to the absolute path of directory which you allow to be used.

Type identifier

Use the following identifier in step "type" field: roboflow_core/local_file_sink@v1to add the block as as step in your workflow.

Properties

Name	Type	Description	Refs
name	str	Enter a unique identifier for this step..	❌
file_type	str	Type of the file.	❌
output_mode	str	Decides how to organise the content of the file.	❌
target_directory	str	Target directory.	✅
file_name_prefix	str	File name prefix.	✅
max_entries_per_file	int	Defines how many datapoints can be appended to a single file.	✅

The Refs column marks possibility to parametrise the property with dynamic values available in workflow runtime. See Bindings for more info.

Available Connections

Compatible Blocks

Check what blocks you can connect to Local File Sink in version v1.

• inputs: VLM as Detector, Google Vision OCR, CSV Formatter, OpenAI, LMM For Classification, Twilio SMS Notification, VLM as Classifier, Single-Label Classification Model, Roboflow Dataset Upload, Roboflow Dataset Upload, Model Monitoring Inference Aggregator, Webhook Sink, OCR Model, Florence-2 Model, EasyOCR, Multi-Label Classification Model, Florence-2 Model, Slack Notification, Clip Comparison, Instance Segmentation Model, OpenAI, Keypoint Detection Model, Google Gemini, Anthropic Claude, LMM, Email Notification, Google Gemini, Llama 3.2 Vision, CogVLM, Email Notification, Local File Sink, OpenAI, Roboflow Custom Metadata, Stitch OCR Detections, Object Detection Model
• outputs: Google Vision OCR, SAM 3, Image Preprocessing, LMM For Classification, Ellipse Visualization, Triangle Visualization, QR Code Generator, Background Color Visualization, Model Monitoring Inference Aggregator, Segment Anything 2 Model, Template Matching, Distance Measurement, Dot Visualization, Halo Visualization, Slack Notification, Color Visualization, Llama 3.2 Vision, Line Counter, Size Measurement, Email Notification, Corner Visualization, Mask Visualization, Time in Zone, Roboflow Custom Metadata, Stability AI Outpainting, Time in Zone, Crop Visualization, Perspective Correction, Single-Label Classification Model, Contrast Equalization, Polygon Zone Visualization, CLIP Embedding Model, Bounding Box Visualization, Icon Visualization, Image Blur, Time in Zone, Path Deviation, Anthropic Claude, Multi-Label Classification Model, Dynamic Crop, Path Deviation, Detections Consensus, Model Comparison Visualization, Cache Get, Local File Sink, Classification Label Visualization, Circle Visualization, Stability AI Inpainting, Moondream2, Florence-2 Model, Morphological Transformation, Reference Path Visualization, Gaze Detection, SIFT Comparison, Polygon Visualization, Florence-2 Model, Clip Comparison, Perception Encoder Embedding Model, Instance Segmentation Model, OpenAI, PTZ Tracking (ONVIF).md), Line Counter, Keypoint Detection Model, Object Detection Model, Google Gemini, Label Visualization, Email Notification, Trace Visualization, Dynamic Zone, YOLO-World Model, OpenAI, CogVLM, Stitch OCR Detections, Detections Stitch, Cache Set, Blur Visualization, Single-Label Classification Model, OpenAI, Detections Classes Replacement, Twilio SMS Notification, Seg Preview, Roboflow Dataset Upload, Roboflow Dataset Upload, Stability AI Image Generation, Webhook Sink, Line Counter Visualization, Instance Segmentation Model, Multi-Label Classification Model, Pixelate Visualization, Image Threshold, Keypoint Detection Model, LMM, Google Gemini, Pixel Color Count, Keypoint Visualization, Object Detection Model

Input and Output Bindings

The available connections depend on its binding kinds. Check what binding kinds Local File Sink in version v1 has.

Bindings

• input

  • content (string): Content of the file to save.
  • target_directory (string): Target directory.
  • file_name_prefix (string): File name prefix.
  • max_entries_per_file (string): Defines how many datapoints can be appended to a single file.

• output

  • error_status (boolean): Boolean flag.
  • message (string): String value.

Example JSON definition of step Local File Sink in version v1

{
    "name": "<your_step_name_here>",
    "type": "roboflow_core/local_file_sink@v1",
    "content": "$steps.csv_formatter.csv_content",
    "file_type": "csv",
    "output_mode": "append_log",
    "target_directory": "some/location",
    "file_name_prefix": "my_file",
    "max_entries_per_file": 1024
}