mem0ai/mem0 · error · Error

Databricks vector store requires either workspaceUrl or…

Error message

Databricks vector store requires either workspaceUrl or host.

What it means

extractHostAndWorkspaceUrl() derives the workspace URL used for every Management/Vector Search API call. It first tries config.workspaceUrl, then config.host, and throws if neither is set — the store has no endpoint to talk to.

Solutions

Set host: 'adb-1234567890123456.7.azuredatabricks.net' (or your cloud's equivalent) in the vector store config
Or set workspaceUrl: 'https://adb-....azuredatabricks.net' — both forms are accepted
If config comes from env, ensure the variable is present and fail config loading early when it is missing

Example fix

// before
new Databricks({ accessToken, catalog: 'main', schema: 'default' });

// after
new Databricks({
  host: 'adb-1234567890123456.7.azuredatabricks.net',
  accessToken,
  catalog: 'main',
  schema: 'default',
});

Defensive patterns

Strategy: validation

Validate before calling

if (!cfg.workspaceUrl && !cfg.host) {
  throw new Error('Databricks config requires workspaceUrl or host');
}

Type guard

const hasDatabricksHost = (c: any): boolean => Boolean(c?.workspaceUrl || c?.host);

Prevention

Make host a required field in a zod schema for the config object
Fail startup when DATABRICKS_HOST is unset rather than passing undefined

When it happens

Trigger: Constructing the Databricks vector store with neither workspaceUrl nor host in the config, e.g. both left undefined because the env vars were not read (DATABRICKS_HOST unset).

Common situations: Missing DATABRICKS_HOST / DATABRICKS_URL in the deployment environment; passing the host inside an auth sub-object or wrong key ('serverUrl', 'workspace'); config built from a partial env where the key is optional and skipped.

Related errors

AI-assisted analysis of mem0ai/mem0@001c235229 (2026-08-15). Data as JSON: /api/errors/27168e2abdd229e1. Report an issue: GitHub.

Appendix: source

Thrown at mem0-ts/src/oss/src/vector_stores/databricks.ts:116

  host: string;
  workspaceUrl: string;
} {
  if (config.workspaceUrl) {
    const url = new URL(config.workspaceUrl);
    return {
      host: url.host,
      workspaceUrl: `${url.protocol}//${url.host}`,
    };
  }

  if (config.host) {
    return {
      host: config.host,
      workspaceUrl: `https://${config.host}`,
    };
  }

  throw new Error(
    "Databricks vector store requires either workspaceUrl or host.",
  );
}

// `Memory.search()` hands keywordSearch() an already-lemmatized query, so BM25 has to score
// against lemmatized text. Same contract as baidu.ts / pgvector.ts.
function lemmatizedText(payload: Record<string, any>): string {
  const data = typeof payload.data === "string" ? payload.data : "";
  return typeof payload.textLemmatized === "string" &&
    payload.textLemmatized.length > 0
    ? payload.textLemmatized
    : data;
}

function formatSqlValue(value: any): string {
  if (value === null || value === undefined) {
    return "NULL";
  }

View on GitHub (pinned to 001c235229)