Unverified50% confidenceCautionExact time
By default, LangExtract silently swallows parsing errors for each chunk: when the model output is wrapped in explanatory text and fences, or when JSON parsing fails because it was truncated due to finish_reason being length, that chunk is discarded and only a log is printed, returning zero extraction results without throwing an exception.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/19/2026
First Seen
Valid until: 1/17/2027
Sources
Related Entities
Related Claims
UnverifiedOnce final JSON parsing fails, subsequent semantic checks are skipped entirely, which is the technical root cause of the 'task completed yet zero score' phenomenon.71% similarUnverifiedllm-retry's with_failover feature silently retries 400 errors (such as prompt too long or JSON format errors) to the fallback model, masking the error and causing quality degradation that no one notices.70% similarUnverifiedIf the logs already carry structured failure records (JSON events containing test IDs, messages, and stack traces), then the reducer is a redundant ritual and this step should be skipped.69% similarUnverifiedThe most common model failure is returning something that is 'almost JSON,' such as extra commas, markdown code fences, verbose preambles, or writing the number 3 as the word 'three,' with the AI node reporting success while the workflow crashes only several steps later when parsing fails.68% similarUnverifiedWhen an LLM node returns malformed JSON, the tool contract specifies not to retry indefinitely, but instead to log the error, skip that record, and mark the number of skipped records in the summary.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1066764API
curl https://kongchang.com/api/v1/knowledge/claims/1066764MCP
get_claim(id=1066764)