zylon-ai/private-gpt · error · InvalidFileError

zpgt.ingest.parsing_failure.error

zpgt.ingest.parsing_failure.error

Error message

zpgt.ingest.parsing_failure.error

What it means

Raised as InvalidFileError with code zpgt.ingest.parsing_failure.error when the primary document reader raised ExtractionUnsuccessfulError and the vision-model fallback reader also failed during extraction. The original reader error and the vision error are both chained as __cause__. It means both the text-extraction path and the OCR/vision path could not parse the file.

Source

Thrown at private_gpt/components/ingest/parse_component.py:121

            )
        except ExtractionUnsuccessfulError as e:
            logger.warning("Extraction unsuccessful for %s: %s", file_info.file_name, e)
            try:
                vision_nodes = self._extract_with_vision_fallback(
                    converted_file,
                    file_metadata,
                    notification=notification,
                    warnings=warnings,
                )
            except Exception as vision_error:
                # Vision reader was available but failed during extraction
                logger.error(
                    "Vision reader fallback failed for %s: %s",
                    file_info.file_name,
                    vision_error,
                    exc_info=True,
                )
                raise InvalidFileError(
                    errors=[IngestionParseErrors.PARSING_FAILURE],
                    warnings=warnings,
                ) from vision_error

            if vision_nodes:
                nodes = vision_nodes
            else:
                raise InvalidFileError(
                    errors=[IngestionParseErrors.PARSING_FAILURE],
                    warnings=warnings,
                ) from e
        except RuntimeError as e:
            raise InvalidFileError(errors=[IngestionParseErrors.PARSING_FAILURE]) from e
        except Exception as e:
            logger.error("Error loading file: %s", e, exc_info=True)

            converted_fallback = convert_unsupported_file_as_fallback(file_info)
            if converted_fallback:

View on GitHub (pinned to 4a030776a3)

Solutions

1. Check the chained exceptions in logs (logger.error prints vision_error with exc_info=True) to identify the underlying vision API failure and fix that (API key, endpoint, network).
2. Verify the document opens in a normal viewer; if corrupted, re-export or repair the source file before ingest.
3. Test the vision reader directly on a simple image to confirm the multimodal backend works.
4. If the file is text-based but malformed, try converting it to PDF via convert_unsupported_file before ingest.

try:
    result = ingest_component.parse_file(file_data, file_metadata)
except InvalidFileError as e:
    if "zpgt.ingest.parsing_failure.error" in e.errors:
        # inspect e.__cause__ chain: vision error vs original extraction error
        logger.warning("parse failed: %s", e.__cause__)
        return reject_upload("unsupported_or_corrupt_document")

When it happens

Trigger: Ingesting a scanned/corrupted PDF or image-heavy document where the standard reader throws ExtractionUnsuccessfulError, _extract_with_vision_fallback is configured (vision LLM available), but the vision API call itself errors (network failure, invalid API key, model timeout, unsupported image payload).

Common situations: Vision LLM credentials missing or expired; vision model endpoint unreachable from the worker; corrupt PDFs that defeat both parsers; oversized pages rejected by the vision API; local deployment where multimodal extra was never installed.

Related errors

• zpgt.ingest.no_valid_files.error
• Error reading delimited file: {e}
• Invalid system item in list (dict): {item}
• Invalid system item in list: {item}
• Invalid system specification: {system}

AI-assisted analysis of zylon-ai/private-gpt@4a030776a3 (2026-08-15). Data as JSON: /api/errors/61c86385ab6c2553. Report an issue: GitHub.