ruvnet/ruflo · error · Error

Invalid embedding value at index

Error message

Invalid embedding value at index ${i}: expected finite number, got ${typeof embedding[i]}

What it means

formatEmbedding() in the ruvector import command validates every element of each embedding array is a finite number before interpolating the array into a PostgreSQL vector literal '[...]'::ruvector(N). null, strings, NaN, or Infinity entries throw with the offending index — this is SQL-injection hardening for JSON-sourced data (CRIT class fix).

Solutions

1. Preprocess the dataset: drop or impute non-numeric elements so every value is a finite number
2. In Python, re-export with json.dumps(data, allow_nan=False) and convert tensors via .tolist() on float32
3. Verify programmatically before import: rows.every(v => typeof v === 'number' && Number.isFinite(v))

Example fix

// before
{"id":"d1","embedding":[0.1,null,"0.3"]}

// after
{"id":"d1","embedding":[0.1,0.0,0.3]}

for (const row of rows) {
  if (!isFiniteVector(row.embedding)) {
    row.embedding = row.embedding.map((v) => (typeof v === 'number' && Number.isFinite(v)) ? v : 0);
  }
}
// only then run the import

function isFiniteVector(v: unknown): v is number[] {
  return Array.isArray(v) && v.every((x) => typeof x === 'number' && Number.isFinite(x));
}

try {
  await importRows(rows);
} catch (err) {
  if (err instanceof Error && /Invalid embedding value at index \d+/.test(err.message)) {
    // sanitize or drop the offending row, then retry the batch
  } else throw err;
}

When it happens

Trigger: Importing a JSON/JSONL dataset whose embedding arrays contain null (Python None), "0.25" as a string, or literal NaN/Infinity written by a non-strict serializer; also mixed-shape rows from truncated exports.

Common situations: Exporting embeddings from Python with missing values; json.dumps with allow_nan=True (default) emitting NaN; schema drift where the producer changed element types between runs.

Related errors

• Invalid schema name: " ". Must contain only letters…
• Invalid timestamp format
• Schema name must not be empty
• each record requires a non-empty numeric vector
• Embedding must be Float32Array of length

AI-assisted analysis of ruvnet/ruflo@fa13ee4ad6 (2026-08-18). Data as JSON: /api/errors/43e182666297a5ca. Report an issue: GitHub.

Appendix: source

Thrown at v3/@claude-flow/cli/src/commands/ruvector/import.ts:52

 */
interface ImportStats {
  total: number;
  imported: number;
  skipped: number;
  errors: number;
  withEmbeddings: number;
  byNamespace: Record<string, number>;
}

/**
 * Format a ruvector embedding array for PostgreSQL
 * Validates each element is a finite number to prevent SQL injection via crafted arrays.
 */
function formatEmbedding(embedding: number[], dimensions: number = 384): string {
  // Validate every element is a finite number (prevents SQL injection via crafted JSON)
  for (let i = 0; i < embedding.length; i++) {
    if (typeof embedding[i] !== 'number' || !Number.isFinite(embedding[i])) {
      throw new Error(`Invalid embedding value at index ${i}: expected finite number, got ${typeof embedding[i]}`);
    }
  }

  // Ensure correct dimensions by padding or truncating
  const padded = [...embedding];
  while (padded.length < dimensions) {
    padded.push(0);
  }
  if (padded.length > dimensions) {
    padded.length = dimensions;
  }
  return `'[${padded.join(',')}]'::ruvector(${dimensions})`;
}

/**
 * Escape string for PostgreSQL
 */
function escapeString(str: string): string {

View on GitHub (pinned to fa13ee4ad6)