/help

Synapse Export Help

Export annotation data, ground truth datasets, and task data from Synapse projects. Supports multiple output formats (JSON, COCO, YOLO, Pascal VOC) and both script-based and plugin-based execution.

Available Commands

Command	Description
/synapse-export:help	Show this help message
/synapse-export:export	Export data from a Synapse project
/synapse-export:export-status	Check status and logs of an export job

Skills (Auto-Activated)

The following skills automatically activate based on conversation context:

Skill	Triggers
export-workflow	"export", "download annotations", "ground truth", "COCO format", "YOLO format", "Pascal VOC", "export assignments", "export tasks", "download labels"

Agents (Autonomous)

Agent	Purpose
export-assistant	Full export orchestrator — analyze annotations, convert formats, execute export

Prerequisites

Requirement	Minimum Version	Check Command
Python	3.12+	python3 --version
synapse-sdk	>= 2026.1.39	synapse --version
Synapse authentication	—	synapse doctor

Environment Setup

# Try the current shell first (venv may already be activated)
synapse --version

# If not found, look for a venv in cwd and activate it
ls -d *venv* .venv 2>/dev/null
source .venv/bin/activate  # or whichever venv dir was found

# If no venv exists, install manually
uv pip install "synapse-sdk[all]>=2026.1.39"

# Authenticate
synapse login

# Verify
synapse doctor

Quick Start

1. Export annotations: /synapse-export:export --project 42 --format coco
2. Export ground truth: /synapse-export:export --target ground_truth --gt-version 15 --format yolo
3. Check job status: /synapse-export:export-status <job-id>

Examples

Export project annotations as COCO

/synapse-export:export --project 42 --target assignment --format coco

Export ground truth as YOLO with train/val split

/synapse-export:export --target ground_truth --gt-version 15 --format yolo --split 0.8/0.2

Export raw JSON with original files

/synapse-export:export --project 42 --format json --save-original-files

Interactive export (no arguments)

/synapse-export:export

Export Targets

Target	Source	Required	Use Case
assignment	Labeler/reviewer work	Project ID	Export labeled annotations
ground_truth	Curated GT datasets	GT version ID	Export training/validation data
task	Task definitions + annotations	Project ID	Export task metadata

Output Formats

Format	Best For	Annotation Types
JSON (Synapse)	Full fidelity export, backup	All types
COCO	Object detection, segmentation	bbox, polygon, keypoint
YOLO	Real-time detection models	bbox only
Pascal VOC	Classic detection benchmarks	bbox only
CSV	Spreadsheet analysis	Classification, simple attributes

How It Works

1. Select target — Choose what to export (assignments, ground truth, tasks)
2. Analyze data — Sample annotations to understand types and classes
3. Choose format — Pick output format based on annotation types
4. Plan export — Show summary with item count, format, and output structure
5. Execute — Write export script and submit via synapse script submit
6. Report — Display results with output file details

Getting Help

• Ask about export formats: "What format should I use for YOLO training?"
• Ask about filtering: "Can I export only reviewed assignments?"
• Ask about splitting: "How do I create train/val/test splits?"
• Ask about large exports: "How do I export 50,000 annotations efficiently?"
• Ask about custom formats: "I need a custom CSV with specific columns"

The relevant skills will automatically load to provide detailed guidance.