mem0ai/mem0 · error · Error

${label} dimension mismatch. Expected ${this.dimension}, got

Error message

${label} dimension mismatch. Expected ${this.dimension}, got ${vector.length}

What it means

assertVectorDimension() throws when a vector passed for insert/update does not match this.dimension, the configured embedding dimension. Databricks Vector Search indexes have a fixed embedding column width; a mismatched vector would be rejected server-side, so the provider validates client-side first with a clearer message.

Source

Thrown at mem0-ts/src/oss/src/vector_stores/databricks.ts:1456

    return {};
  }

  private extractSessionValues(payload: Record<string, any>): {
    user_id: any;
    agent_id: any;
    run_id: any;
  } {
    return {
      user_id: payload.user_id,
      agent_id: payload.agent_id,
      run_id: payload.run_id,
    };
  }

  private assertVectorDimension(vector: number[], label: string): void {
    if (vector.length !== this.dimension) {
      throw new Error(
        `${label} dimension mismatch. Expected ${this.dimension}, got ${vector.length}`,
      );
    }
    for (const value of vector) {
      if (!Number.isFinite(value)) {
        throw new Error(
          `${label} values must be finite numbers for Databricks vector search.`,
        );
      }
    }
  }

  private matchFieldCondition(
    vector: DatabricksVector,
    key: string,
    value: any,
  ): boolean {
    const fieldValue = key === "memory_id" ? vector.id : vector.payload[key];

View on GitHub (pinned to 001c235229)

Solutions

Make the embedding model used by Memory match the dimension the Databricks index was created with, and pass that same dimension in the store config.
If you intentionally changed embedding models, create a new Databricks Vector Search index with the new dimension and point config at it.
Log vector.length at the call site to identify which embedder produced the mismatched vector.
Verify config.dimension matches the index's embedding column length in Databricks.

Example fix

// before
const store = new DatabricksDB({ ...opts, dimension: 1536 });
const memory = new Memory({ vectorStore: store, embedder: new OllamaEmbedder() }); // 768-dim

// after
const memory = new Memory({
  vectorStore: new DatabricksDB({ ...opts, dimension: 768 }), // matches embedder
  embedder: new OllamaEmbedder(),
});

Defensive patterns

Strategy: validation

Validate before calling

const dim = await embedder.embed('probe');
if (dim.embedding.length !== storeConfig.dimension) {
  throw new Error(`Embedder emits ${dim.embedding.length}-d vectors but store configured for ${storeConfig.dimension}`);
}

Type guard

const isDimensionOk = (v: number[], expected: number) =>
  Array.isArray(v) && v.length === expected;

Try / catch

try {
  await store.insert(vectors, ids, payloads);
} catch (e) {
  if (e instanceof Error && /dimension mismatch/i.test(e.message)) {
    // align embedder and index dimension, recreate index if the model changed
  }
  throw e;
}

Prevention

Derive config.dimension from a probe embedding at startup instead of hardcoding it.
Use a single embedding model across memory add/search/store paths.
Recreate the Databricks index when switching embedding models.

When it happens

Trigger: Calling insert() or update() with vectors produced by a different embedding model than the one whose dimension the store was configured with (e.g. 1536-dim OpenAI vectors sent to a 768-dim index), or mixing embedding providers between memory add and store config.

Common situations: Switching the embedding provider (e.g. from OpenAI text-embedding-3-small to a local 384-dim model) without recreating the Databricks index; sharing one store across two memory instances with different embedders; a custom embedding function returning padded/truncated vectors.

Related errors

AI-assisted analysis of mem0ai/mem0@001c235229 (2026-08-15). Data as JSON: /api/errors/c0da84ff33faa537. Report an issue: GitHub.