Embedding Vector Validator

Overview

You send one embedding vector of 1 to 8,192 finite numbers. We report dimensions, compensated L1 and overflow-resistant L2 norms, min, max, zero count and diagnostics, and we never echo the vector. expectedDimensions, when sent, must match the length. normalization L2_UNIT holds the L2 norm to unitTolerance (default 0.000001). zeroVector REJECT, the default, fails a all-zero vector; ALLOW keeps it. We do not generate embeddings, compare two vectors, or change normalization.

An index that assumed 1,536 dimensions or a unit vector silently stores a zero or a short row and poisons retrieval.

When to use it

Scope

Non-goals

• Does not generate embeddings or call a model
• Does not compare two vectors or judge semantic quality
• Does not re-normalize the vector or return a unit copy

Properties

Capabilities

Cost & billing

Flat rate per invocation

File handling

How we handle your data

Input schema

Output schema

API contract

curl -sS -X POST "$BASE/v1/execute" -H "Authorization: Bearer $KEY" -H 'Content-Type: application/json' -d '{"skillKey":"ai.embedding_vector_validate","input":{"vector":[0.6,0.8],"normalization":"L2_UNIT"}}'

import { Squillify } from "@squillify/sdk";

const sq = new Squillify({ apiKey: process.env.SQUILLIFY_API_KEY });

const { output } = await sq.execute("ai.embedding_vector_validate", {"vector":[0.6,0.8],"normalization":"L2_UNIT"});

Errors

Other limits

Limitations

• Does not generate embeddings or call a model
• Does not compare two vectors or judge semantic quality
• Does not re-normalize the vector or return a unit copy
• Producing embeddings
• Nearest-neighbour search

How to call

POST /v1/execute (requires authentication)

{
  "skillKey": "ai.embedding_vector_validate",
  "input": {
    "vector": [
      0.6,
      0.8
    ],
    "normalization": "L2_UNIT"
  }
}

Examples

{
  "vector": [
    0.6,
    0.8
  ],
  "normalization": "L2_UNIT"
}

{
  "valid": true,
  "normalization": "L2_UNIT",
  "min": 0.6,
  "nonFiniteCount": 0,
  "max": 0.8,
  "zeroVector": "REJECT",
  "zeroCount": 0,
  "l1Norm": 1.4,
  "l2Norm": 1,
  "diagnostics": [],
  "dimensions": 2,
  "unitTolerance": 0.000001
}

{
  "vector": [
    0,
    0
  ],
  "expectedDimensions": 3
}

{
  "valid": false,
  "normalization": "NONE",
  "min": 0,
  "nonFiniteCount": 0,
  "max": 0,
  "zeroVector": "REJECT",
  "zeroCount": 2,
  "l1Norm": 0,
  "l2Norm": 0,
  "diagnostics": [
    {
      "code": "DIMENSION_MISMATCH",
      "severity": "ERROR",
      "message": "The vector length does not match expectedDimensions.",
      "path": "/vector",
      "source": "squillify",
      "details": {}
    },
    {
      "code": "ZERO_VECTOR",
      "severity": "ERROR",
      "message": "A zero vector is rejected under the published zeroVector policy.",
      "path": "/vector",
      "source": "squillify",
      "details": {}
    }
  ],
  "dimensions": 2,
  "unitTolerance": 0.000001
}

Recommended use cases

• Reject a zero vector before you store it as an embedding
• Confirm a provider vector has the dimensions your index expects

Avoid for

• Producing embeddings
• Nearest-neighbour search

Related skills

Composable with

Safety & privacy

• Runs without calling any external service.
• Your payloads are never stored permanently.
• Telemetry captures metadata only — never your payload contents.
• We never train models on your payloads.
• No human reviews your payloads.

Tags