Export API
Export taste profiles for offline use, backup, or integration with external systems. Choose between full exports or minimal inference-only data.
Export Formats
Full Export (json)
Complete profile data including all branches, versions, exemplars, comparisons, and computed snapshots. Ideal for backup or migration.
?format=jsonMinimal Export (minimal)
Only the essential data needed for inference: latent scores, embedding centroid, and prompt summary. Smaller payload for edge deployment.
?format=minimal/v1/profiles/:id/exportExport a taste profile in JSON format. By default, exports the full profile data.
Query Parameters
| Parameter | Type | Description |
|---|---|---|
| format | string | Optional. Either "json" (default) or "minimal" |
curl https://api.commandAGI.com/v1/profiles/prof_abc123/export \
-H "Authorization: Bearer YOUR_API_KEY"
# Or explicitly:
curl "https://api.commandAGI.com/v1/profiles/prof_abc123/export?format=json" \
-H "Authorization: Bearer YOUR_API_KEY"{
"id": "prof_abc123",
"projectId": "proj_xyz789",
"name": "cinematic-look",
"seed": "Dark, moody cinematography",
"profileData": {
"branches": [
{
"id": "main",
"name": "main",
"description": "Main branch",
"headVersionId": "v_456",
"createdAt": "2024-01-15T10:30:00Z"
}
],
"versions": [
{
"id": "v_456",
"branchId": "main",
"message": "Added 20 comparisons",
"snapshot": {...},
"createdAt": "2024-01-20T14:22:00Z"
}
],
"currentBranchId": "main",
"budget": {
"questions": { "used": 15, "total": 20 },
"labels": { "used": 42, "total": 50 },
"comparisons": { "used": 87, "total": 100 }
}
},
"effectiveSnapshot": {
"constraints": [
{
"id": "c_1",
"category": "mood",
"question": "What mood should the image evoke?",
"answer": "Mysterious and contemplative"
}
],
"exemplars": [
{
"id": "e_1",
"label": "good",
"imageData": "https://cdn.commandAGI.com/...",
"source": "upload"
}
],
"comparisons": [
{
"id": "cmp_1",
"frameA": "frm_1",
"frameB": "frm_2",
"winner": "frm_1"
}
],
"latentScores": {
"frm_1": 0.823,
"frm_2": 0.456,
"frm_3": 0.912
},
"promptSummary": "Dark, cinematic imagery with high contrast...",
"embeddingCentroid": [0.12, -0.34, 0.56, ...]
},
"metadata": {
"createdAt": "2024-01-15T10:30:00Z",
"updatedAt": "2024-01-20T14:22:00Z",
"exportedAt": "2024-01-21T09:00:00Z",
"version": "1.0"
}
}Minimal Export
The minimal format includes only the data required for inference, significantly reducing payload size. Ideal for edge deployment or embedding in applications.
curl "https://api.commandAGI.com/v1/profiles/prof_abc123/export?format=minimal" \
-H "Authorization: Bearer YOUR_API_KEY"{
"id": "prof_abc123",
"name": "cinematic-look",
"seed": "Dark, moody cinematography",
"snapshot": {
"latentScores": {
"frm_1": 0.823,
"frm_2": 0.456,
"frm_3": 0.912
},
"embeddingCentroid": [0.12, -0.34, 0.56, ...],
"promptSummary": "Dark, cinematic imagery with high contrast...",
"exemplarCount": 42,
"comparisonCount": 87,
"constraintCount": 15
},
"exportedAt": "2024-01-21T09:00:00Z"
}Response Fields
Full Export
| Field | Description |
|---|---|
| profileData | Complete profile structure with branches, versions, and budget |
| effectiveSnapshot | All constraints, exemplars, comparisons, and computed data |
| metadata | Export metadata including timestamps and version |
Minimal Export
| Field | Description |
|---|---|
| snapshot.latentScores | Bradley-Terry rankings for all exemplars |
| snapshot.embeddingCentroid | Average embedding vector for similarity scoring |
| snapshot.promptSummary | Natural language description of the taste |
| snapshot.*Count | Counts of exemplars, comparisons, and constraints |
Use Cases
Offline Evaluation
Export the minimal format and use the embedding centroid for local similarity calculations without API calls.
Backup and Versioning
Use full exports to create point-in-time backups of your taste profiles. Store in version control alongside your codebase.
Edge Deployment
Embed minimal exports in edge functions for low-latency taste scoring without round-trips to the API.
Cross-Platform Integration
Export profiles to integrate with external ML pipelines, analytics tools, or custom applications.
Implementation Tips
- Cache exports: Export data changes only when you commit new versions. Cache the export and refresh on version changes.
- Use minimal for inference: The full export can be large. Use minimal format when you only need to score content.
- Store the centroid: The embedding centroid is the most valuable data for offline scoring. A simple cosine similarity check provides good results.