zylon-ai/private-gpt · error · InvalidFileError

zpgt.ingest.no_valid_nodes.error

zpgt.ingest.no_valid_nodes.error

Error message

zpgt.ingest.no_valid_nodes.error

What it means

InvalidFileError raised when extended_index.insert()/ainsert() returns an empty result — the document parsed into nodes, but the vector store inserted none of them. Carries the IngestionLoadErrors.NO_VALID_NODES code (zpgt.ingest.no_valid_nodes.error) and the warnings list. At this point the index status is not advanced to POPULATED, so the collection stays empty for this artifact.

Source

Thrown at private_gpt/components/ingest/ingest_component.py:279

            )

            # 1. Delete previous nodes, to avoid duplicates
            self.node_store_component.delete_filtered_nodes(
                collection=collection,
                artifacts=[artifact],
            )

            # 2. Insert nodes
            inserted_nodes: Sequence[BaseNode] = []
            if use_async:
                inserted_nodes = asyncio.run(
                    extended_index.ainsert(nodes, notify=notify_publisher)
                )
            else:
                inserted_nodes = extended_index.insert(nodes, notify=notify_publisher)

            if not inserted_nodes:
                raise InvalidFileError(
                    errors=[IngestionLoadErrors.NO_VALID_NODES], warnings=warnings
                )

            index.summary = ArtifactIndexStatus.POPULATED.value
            index.set_index_id(index_id)
            index.storage_context.persist(persist_dir=local_data_path / collection)
            logger.info("Finished loading index %s with %d nodes", index_id, len(nodes))

View on GitHub (pinned to 4a030776a3)

Solutions

  1. Check the warnings attached to the error and the parse result — they usually name why nodes were dropped
  2. Verify the document actually contains extractable text (run OCR for scanned PDFs)
  3. Confirm the embedding model's dimensions match the target collection/schema
  4. Log the nodes passed to insert to see whether they were empty before reaching the store

Example fix

# before: insert returns [] -> InvalidFileError(NO_VALID_NODES)
nodes = result.nodes
inserted = extended_index.insert(nodes, notify=notify_publisher)

# after
clean = [n for n in nodes if (n.get_content() or '').strip()]
if not clean:
    raise ValueError('document has no extractable text')
inserted = extended_index.insert(clean, notify=notify_publisher)
Defensive patterns

Strategy: try-catch

Validate before calling

contentful = [n for n in nodes if (n.get_content() or '').strip()]
if not contentful:
    raise ValueError('no nodes with non-empty text; document likely lacks extractable text')

Try / catch

try:
    component._ingest(file, collection)
except InvalidFileError as e:
    if IngestionLoadErrors.NO_VALID_NODES in (e.errors or []):
        return error_response(422, 'no valid nodes produced', warnings=e.warnings)
    raise

Prevention

When it happens

Trigger: Running _ingest with use_async true or false where insert(nodes, notify=...) returns [] — e.g. all nodes filtered out by the store's transforms/embedding rules, empty text nodes, unsupported node types, or a store silently rejecting the batch.

Common situations: Documents whose extracted text is empty or whitespace-only (scanned PDFs without OCR); node transforms dropping everything; dimension mismatch between embedding output and the vector store collection; store insert hooks filtering by metadata; silent failures in a custom ExtendedIndex.

Related errors

AI-assisted analysis of zylon-ai/private-gpt@4a030776a3 (2026-08-15). Data as JSON: /api/errors/00ff53ebdf8efbc5. Report an issue: GitHub.