Why do code evaluators fail with "Code eval dispatcher is not configured"?
Code evaluators run your evaluation code outside the main Langfuse process. On self-hosted Langfuse you choose where that code runs with LANGFUSE_CODE_EVAL_DISPATCHER. In production builds this variable has no default, so testing or running a code evaluator fails with Code eval dispatcher is not configured until you set it.
Set the variable on both the web and worker containers:
| Value | Runs | Notes |
|---|---|---|
aws-lambda | Python and TypeScript | Executes code in AWS Lambda functions. Configure the function names and IAM permissions as described in AWS Lambda dispatcher. |
external | Python and TypeScript, or the subset in LANGFUSE_CODE_EVAL_EXTERNAL_LANGUAGES | Sends code to an HTTP executor you operate, with a signed request. Set the endpoint and signing secret as described in External HTTP dispatcher. |
insecure-local | TypeScript/JavaScript only | Runs code inside the worker process. Not a sandbox; use only when every user who can create evaluators is trusted. See Local insecure dispatcher. |
Restart web and worker after changing the variable. Code evaluators are available in open-source self-hosted Langfuse; no license key is required.
Was this page helpful?
Last updated on
