mem0ai/mem0 · error · Error

${context} dimension mismatch. Expected ${this.dimension}, g

Error message

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

What it means

Every vector written to or queried against the S3 Vectors index must have exactly this.dimension components (fixed by the embedding model / index creation). assertVectorDimension throws with the failing context (e.g. insert, query) when a vector's length differs, protecting the index from corrupt data and AWS-side rejects.

Source

Thrown at mem0-ts/src/oss/src/vector_stores/s3_vectors.ts:862

  private normalizeScore(
    distance?: number,
    distanceMetric: "cosine" | "euclidean" = this.distanceMetric,
  ): number | undefined {
    if (distance === undefined || distance === null) {
      return undefined;
    }
    if (!Number.isFinite(distance)) {
      return undefined;
    }
    if (distanceMetric === "euclidean") {
      return 1 / (1 + distance);
    }
    return Math.max(0, Math.min(1, 1 - distance));
  }

  private assertVectorDimension(vector: number[], context: string): void {
    if (vector.length !== this.dimension) {
      throw new Error(
        `${context} dimension mismatch. Expected ${this.dimension}, got ${vector.length}`,
      );
    }
  }

  private assertBatchDimensions(vectors: number[][], context: string): void {
    for (const vector of vectors) {
      this.assertVectorDimension(vector, context);
    }
  }

  private isNotFound(error: any): boolean {
    return error?.name === "NotFoundException";
  }

  private isConflict(error: any): boolean {
    return error?.name === "ConflictException";
  }

View on GitHub (pinned to 001c235229)

Solutions

Align the embedding model and embeddingModelDims config with the dimension the index was created with, then recreate the collection if the model changed.
If you supply vectors yourself, verify vector.length === configured dimension before calling add/search.
Re-index existing memories with the new embedding model into a fresh collection.

Example fix

// before
const memory = new Memory({ vectorStore: { provider: 's3_vectors', config: { embeddingModelDims: 1536 } }, embedder: new OpenAIEmbedding({ model: 'text-embedding-3-large' }) }); // 3072-dim model

// after
const memory = new Memory({ vectorStore: { provider: 's3_vectors', config: { embeddingModelDims: 3072 } }, embedder: new OpenAIEmbedding({ model: 'text-embedding-3-large' }) });

Defensive patterns

Strategy: type-guard

Validate before calling

const expected = storeConfig.embeddingModelDims;
if (vector.length !== expected) throw new Error(`Vector has ${vector.length} dims, index expects ${expected}; re-embed or recreate index`);

Type guard

function isCorrectDimension(vector: number[], dims: number): vector is number[] & { length: dims } {
  return Array.isArray(vector) && vector.length === dims;
}

Try / catch

try { await memory.add(text, { embeddingVector }); } catch (e) { if (e instanceof Error && e.message.includes('dimension mismatch')) { /* re-embed with the configured model or recreate index */ } else throw e; }

Prevention

Pin embeddingModelDims to the model's actual output
Recreate the collection when switching embedders
Add a startup assert comparing a sample embedding's length to the config

When it happens

Trigger: Calling add/search with a custom embedding whose dimension differs from the index; switching embedding models (e.g. text-embedding-3-small 1536 -> bge 768) without recreating the S3 Vectors index/collection; mixing manually supplied vectors with model-generated ones.

Common situations: Changing the embedder config after data was already indexed; using embeddingModelDims that doesn't match the actual model output; passing truncated or padded vectors from a custom pipeline.

Related errors

AI-assisted analysis of mem0ai/mem0@001c235229 (2026-08-15). Data as JSON: /api/errors/3806f9dc397d782d. Report an issue: GitHub.