The provider's content filter stopped the call: it blocked the prompt, or stopped the reply, before or after some text (Gemini's promptFeedback.blockReason, or a finishReason such as SAFETY, RECITATION, PROHIBITED_CONTENT, BLOCKLIST or SPII). A reply the filter cut short is not a finished answer; a stream has already delivered the text before the stop, and the message says how much there was.
The provider is healthy, so a ReliableClient's circuit breaker counts this as neither a success nor a failure. As a NonRecoverableError, sending the same request again would be filtered the same way: the conversation must change before a retry.
Value parameters
- promptBlocked
-
whether the prompt itself was blocked (
blockReason), rather than the reply (finishReason) - provider
-
the client's provider name, e.g.
Gemini - reason
-
the provider's reason, as it named it, e.g.
SAFETY
Attributes
- Companion
- object
- Source
- ContentFilteredError.scala
- Graph
-
- Supertypes
-
trait NonRecoverableErrortrait LLMErrortrait Serializabletrait Producttrait Equalsclass Objecttrait Matchableclass Any
