BamlClientFinishReasonError

Thrown when an LLM terminates with a disallowed finish reason.

Configuration

Control allowed finish reasons in your client definition:

client<llm> OpenAIWithFinishReasonError {
  provider openai
  options {
    api_key env.OPENAI_API_KEY
    model "gpt-4"
    max_tokens 10  // Small limit to trigger early stop
    // Throws if model returns any other finish reason
    finish_reason_allow_list ["stop"]
    // Or deny specific reasons
    // finish_reason_deny_list ["length"]
  }
}

Type Definition

class BamlClientFinishReasonError extends Error {
  type: 'BamlClientFinishReasonError'
  message: string
  prompt: string
  raw_output: string
  detailed_message: string
}

Properties

Property	Type	Description
type	'BamlClientFinishReasonError'	Literal type identifier
message	string	Finish reason that caused termination
prompt	string	Original prompt sent to LLM
raw_output	string	Partial output received before termination
detailed_message	string	Complete history of failed attempts (for fallback/retry)

Detection

TypeScript

import { BamlClientFinishReasonError } from '@boundaryml/baml'

try {
  const result = await b.ExtractUser(text)
} catch (error) {
  if (error instanceof BamlClientFinishReasonError) {
    console.log('Finish reason error:', error.message)
    console.log('Partial output:', error.raw_output)
  }
}

Python

from baml_py import BamlClientFinishReasonError

try:
    result = await b.ExtractUser(text)
except BamlClientFinishReasonError as e:
    print(f"Finish reason error: {e}")
    print(f"Partial output: {e.raw_output}")

Common Finish Reasons

Reason	Description
stop	Natural completion
length	Hit max_tokens limit
content_filter	Content blocked by safety filters
tool_calls	Model wants to call a tool

See Also

• Error Overview
• BamlValidationError
• BamlAbortError