Stage files for parsing and indexing.

This intake flow supports PDF, DOCX, and supplier evidence bundles. Metadata is captured up front so the downstream parser and retrieval pipeline stay normalized.

PDF

Drop files here or browse your local archive.

Recommended per batch: up to 20 files, 200 MB combined, one regulation family per batch.

GBT31484_revision_notes.pdf

size pending validation · scanned appendix included

Ready

supplier_thermal_test_report.docx

size pending validation · confidential evidence supplement

Metadata

Regulation type

Version / release

Owning team

Parser backend

Reviewer note

Import queue

Batch progress and validation feedback

Reviewers can monitor preflight checks before the parser job is submitted, then watch queue depth and failure mode without leaving the modal.

2 files validated

Files already validated for metadata completeness

live queue

Current Aliyun parser queue depth populates here

text-embedding-v3

Embedding target applied after semantic block extraction

Preflight validation

Duplicate ID scan, file-type validation, metadata completeness

Passed

Object storage upload

Bucket `upload-files` · transient object names attached to batch

Completed

Aliyun parse submission

Polling every 5s · timeout 900s

Queued

Chunking + embedding

Build semantic blocks, overlapping vector chunks, then 1024-d embeddings

Waiting

Potential issue to resolve

`supplier_thermal_test_report.docx` lacks a formal standard number. The batch can continue, but retrieval quality improves if you assign an evidence relationship before submit.