Failed to obtain the embedding dimensions from the…

Q: What causes "Failed to obtain the embedding dimensions from the embedding model and fall backs to default: ${OPENAI_EMBEDDING_DIMENSION_SIZE}" in spring-projects/spring-ai?

Explicitly set the dimension in the builder (withDimensions / vectorTableName config) instead of relying on auto-detection

Error message

Failed to obtain the embedding dimensions from the embedding model and fall backs to default: ${OPENAI_EMBEDDING_DIMENSION_SIZE}

What it means

MariaDBVectorStore.embeddingDimensions() asks the configured EmbeddingModel for its vector dimension; if that call throws (or returns <= 0) the store logs a warning and falls back to OPENAI_EMBEDDING_DIMENSION_SIZE (1536). The table is then created with the default dimension, which will break inserts for non-OpenAI models whose dimensions differ.

Solutions

Explicitly set the dimension in the builder (withDimensions / vectorTableName config) instead of relying on auto-detection
Verify the embedding model bean is properly configured (API key, base URL) before the store initializes
Check why embeddingModel.dimensions() returns <= 0 for your model and upgrade the model integration
If the fallback already happened, drop/recreate the table with the correct vector column dimension

Example fix

// before
MariaDbVectorStore.builder(jdbcTemplate, embeddingModel).build();

// after
MariaDbVectorStore.builder(jdbcTemplate, embeddingModel)
    .withDimensions(768) // match your embedding model
    .build();

Defensive patterns

Strategy: validation

Validate before calling

int dims = embeddingModel.dimensions();
if (dims <= 0) {
    throw new IllegalStateException("Embedding model did not report a positive dimension; set it explicitly in the builder");
}

Try / catch

try {
    store.afterPropertiesSet();
} catch (Exception e) {
    logger.error("MariaDB vector store init fell back to default dimensions", e);
}

Prevention

Always set withDimensions(n) for non-OpenAI models
Warm up the embedding client before store initialization
Check API key/base URL config in integration tests
Assert the vector column dimension matches model output in schema migration

When it happens

Trigger: Calling afterPropertiesSet / builder / dim while the embeddingModel.dimensions() call throws — e.g. embeddingModel is null, remote embedding endpoint unreachable, API key missing, or a custom model returns 0.

Common situations: Using a non-OpenAI embedding model (e.g. Ollama, Azure, HuggingFace) whose dimensions() call fails or needs network access at startup; missing API key during schema initialization.

Related errors

AI-assisted analysis of spring-projects/spring-ai@98a7beda4f (2026-09-11). Data as JSON: /api/errors/f0338f1446499973. Report an issue: GitHub.

Appendix: source

Thrown at vector-stores/spring-ai-mariadb-store/src/main/java/org/springframework/ai/vectorstore/mariadb/MariaDBVectorStore.java:446

			return this.schemaName + "." + this.vectorTableName;
		}
		return this.vectorTableName;
	}

	int embeddingDimensions() {
		// The manually set dimensions have precedence over the computed one.
		if (this.dimensions > 0) {
			return this.dimensions;
		}

		try {
			int embeddingDimensions = this.embeddingModel.dimensions();
			if (embeddingDimensions > 0) {
				return embeddingDimensions;
			}
		}
		catch (Exception e) {
			logger.warn("Failed to obtain the embedding dimensions from the embedding model and fall backs to"
					+ " default:" + OPENAI_EMBEDDING_DIMENSION_SIZE, e);
		}
		return OPENAI_EMBEDDING_DIMENSION_SIZE;
	}

	@Override
	public VectorStoreObservationContext.Builder createObservationContextBuilder(String operationName) {

		VectorStoreObservationContext.Builder builder = VectorStoreObservationContext
			.builder(VectorStoreProvider.MARIADB.value(), operationName)
			.collectionName(this.vectorTableName)
			.dimensions(this.embeddingDimensions())
			.similarityMetric(getSimilarityMetric());
		if (this.schemaName != null) {
			builder.namespace(this.schemaName);
		}
		return builder;
	}

View on GitHub (pinned to 98a7beda4f)