Skip to content

Hypergraph is more powerful. Transform unstructured text into structured knowledge with LLMs. Graphs, hypergraphs, and spatio-temporal extractions — with one command.

fix: tolerate LLM parse failures during extraction (#45)

dex0shubhamb3e449722802
diff --git a/hyperextract/types/base.py b/hyperextract/types/base.py
index 44f2762..3edb0b1 100644
--- a/hyperextract/types/base.py
+++ b/hyperextract/types/base.py
@@ -257,7 +257,13 @@ class BaseAutoType(ABC, Generic[T]):
if len(text) <= self.chunk_size:
logger.debug("stage=extract_single_chunk chunk_text_preview=%s", text[:200])
- extracted_data = self.data_extractor.invoke({"source_text": text})
+ try:
+ extracted_data = self.data_extractor.invoke({"source_text": text})
+ except Exception as e:
+ # LLM returned unparseable output (e.g. prose instead of JSON).
+ # Log and treat as an empty result instead of crashing the run.
+ logger.warning("stage=extract_single_chunk_failed error=%s", e)
+ extracted_data = None
logger.debug(
"stage=extract_single_chunk_result chunk=0 result_summary=%s",
self._summarize_extracted(extracted_data),
@@ -279,9 +285,22 @@ class BaseAutoType(ABC, Generic[T]):
self.max_workers,
len(inputs),
)
+ # return_exceptions keeps one bad chunk from aborting the whole batch;
+ # failures are logged and nulled (then filtered) below.
extracted_data_list = self.data_extractor.batch(
- inputs, config={"max_concurrency": self.max_workers}
+ inputs,
+ config={"max_concurrency": self.max_workers},
+ return_exceptions=True,
)
+ cleaned = []
+ for i, r in enumerate(extracted_data_list):
+ if isinstance(r, Exception):
+ logger.warning(
+ "stage=chunk_extract_failed chunk_index=%d error=%s", i, r
+ )
+ r = None
+ cleaned.append(r)
+ extracted_data_list = cleaned
logger.debug(
"stage=llm_batch_complete results=%d", len(extracted_data_list)
)