ruvnet/ruflo · error

Unsupported index type: ${options.indexType}

Error message

Unsupported index type: ${options.indexType}

What it means

Thrown by createIndex() when options.indexType is not a key of INDEX_TYPE_SQL. Supported types are 'hnsw', 'ivfflat', 'ivfpq', 'flat', and 'diskann' (the last two being special-cased: flat skips index creation, diskann falls back to HNSW). Any other string yields undefined in the lookup and is rejected as unsupported.

Source

Thrown at v3/@claude-flow/plugins/src/integrations/ruvector/ruvector-bridge.ts:864

    const durationMs = Date.now() - startTime;
    const deleted = result.affectedRows ?? 0;

    return {
      total: ids.length,
      successful: deleted,
      failed: ids.length - deleted,
      durationMs,
      throughput: deleted / (durationMs / 1000),
    };
  }

  /**
   * Create a vector index.
   */
  async createIndex(options: VectorIndexOptions): Promise<void> {
    const indexType = INDEX_TYPE_SQL[options.indexType];
    if (!indexType && options.indexType !== 'flat') {
      throw new Error(`Unsupported index type: ${options.indexType}`);
    }

    const indexName = options.indexName ??
      `idx_${options.tableName}_${options.columnName}_${options.indexType}`;
    const schemaPrefix = this.config.schema ? `${this.escapeIdentifier(this.config.schema)}.` : '';

    if (options.replace) {
      await this.connectionManager.query(
        `DROP INDEX IF EXISTS ${schemaPrefix}${this.escapeIdentifier(indexName)}`
      );
    }

    if (options.indexType === 'flat') {
      return; // No index needed for brute force
    }

    // Build operator class based on metric
    const opClass = this.getOperatorClass(options.metric ?? 'cosine', options.indexType);

View on GitHub (pinned to fa13ee4ad6)

Solutions

1. Use one of: 'hnsw', 'ivfflat', 'ivfpq', 'flat', 'diskann' (lowercase)
2. Type the field as the library's VectorIndexType union so invalid literals fail at compile time
3. Validate/normalize config-sourced values (trim + lowercase + allowlist) before calling createIndex

Example fix

// before
await bridge.createIndex({ tableName: 'vectors', columnName: 'embedding', indexType: 'HNSW' }); // throws

// after
await bridge.createIndex({ tableName: 'vectors', columnName: 'embedding', indexType: 'hnsw' });

const INDEX_TYPES = ['hnsw', 'ivfflat', 'ivfpq', 'flat', 'diskann'] as const;
const indexType = String(cfg.indexType).trim().toLowerCase() as (typeof INDEX_TYPES)[number];
if (!INDEX_TYPES.includes(indexType)) {
  throw new Error(`indexType must be one of: ${INDEX_TYPES.join(', ')}`);
}
await bridge.createIndex({ tableName, columnName, indexType });

const isVectorIndexType = (t: unknown): t is 'hnsw' | 'ivfflat' | 'ivfpq' | 'flat' | 'diskann' =>
  typeof t === 'string' &&
  ['hnsw', 'ivfflat', 'ivfpq', 'flat', 'diskann'].includes(t);

try {
  await bridge.createIndex(opts);
} catch (err) {
  if (err instanceof Error && err.message.startsWith('Unsupported index type')) {
    throw new ConfigurationError(`${opts.indexType} is not a pgvector index type this bridge supports`);
  }
  throw err;
}

When it happens

Trigger: Passing 'ivf', 'lsh', 'HNSW' (wrong case), or a value read straight from user config into createIndex(); forwarding an indexType meant for a different vector DB (e.g. pgvector syntax like 'hnsw_cosine') verbatim.

Common situations: Porting configs from another vector store whose type names differ; config fields typed as plain string instead of the VectorIndexType union; casing drift after a config refactor.

Related errors

• Index ${indexName} not found
• File not found
• Invalid embedding value at index ${i}: expected finite numbe
• Schema name must not be empty
• Schema name too long (${schema.length} chars, max 63): "${sc

AI-assisted analysis of ruvnet/ruflo@fa13ee4ad6 (2026-08-18). Data as JSON: /api/errors/2f797ee49c7da341. Report an issue: GitHub.