EchoStash
Docs

Data Models

Prompt and metadata schemas

PromptEnvelope

The core data structure for every prompt in PLP:

json
{
  "id": "marketing/welcome-email",
  "content": "Hello {{name}}, welcome to {{product}}!",
  "meta": {
    "version": "1.2.0",
    "author": "alice@company.com",
    "description": "Welcome email template",
    "model_config": { "temperature": 0.7, "max_tokens": 1024 }
  }
}
  • id (string, required) — Unique identifier, may include / for namespacing
  • content (string | ContentPart[], required) — Prompt text or multimodal array
  • meta (object, optional) — Version, author, model_config, plus any custom fields

Multimodal Content

When content includes images, it's an array of content parts:

json
{
  "content": [
    { "type": "text", "text": "Analyze this image:" },
    { "type": "image_url", "image_url": { "url": "data:image/png;base64,..." } },
    { "type": "text", "text": "What do you see?" }
  ]
}

Discovery Response

json
{
  "plp_version": "1.1.0",
  "server": "My PLP Server",
  "capabilities": {
    "versioning": true,
    "list": true,
    "contextStore": true,
    "deploy": true,
    "evaluation": true
  }
}

DeployResponse

json
{
  "promptId": "welcome-email",
  "versionNo": 3,
  "environment": "production",
  "deployedAt": "2024-06-15T10:30:00Z"
}

EvalSuiteResult

json
{
  "suiteName": "Quality Check",
  "status": "pass",
  "tests": [{ "name": "...", "status": "pass", "assertions": [...] }],
  "summary": { "total": 3, "passed": 3, "failed": 0, "durationMs": 2450 }
}

ContextStoreAsset

json
{
  "id": 42,
  "assetId": "company-logo",
  "mimeType": "image/png",
  "fileSize": 24576,
  "plpReference": "plp://company-logo",
  "createdAt": "2024-06-15T10:00:00Z"
}