Document Processing
Bunyan prepares uploaded documents for search by reading their content, splitting it into smaller sections, and using the collection's embedding model to make those sections searchable. Processing starts automatically after upload or import.
Check When a Document Is Ready
- Open Document Collections and select the collection.
- Find the document in the list.
- Select Refresh documents to update its Status.
- Wait for Complete, then select the collection or document in the Playground.
You may see intermediate statuses while Bunyan prepares the document:
| Status | Meaning |
|---|---|
| Pending or Uploading | The document is waiting or being received. |
| Processing, Parsing Markdown, Markdown Parsed, or LLM Enhancement | Bunyan is preparing the document's text. These are intermediate stages. |
| Generating Embeddings or Processing Embeddings | Bunyan is preparing the text for search. A percentage may appear. |
| Complete | Processing has finished. The document is ready to use as a source. |
| Error | Processing did not finish successfully. Check the file or import settings before trying again. |
Not every document passes through every displayed stage. Processing time depends on the document and the availability of the models used to process it.
Monitor a Web or Storage Import
The Document Import Progress dialog shows overall progress, files processed, and a failed-file count when applicable. It updates while open. You can close it and the import continues in the background.
An import can finish with some files unsuccessful. Review the counts and the individual document statuses in the collection before assuming that all sources are ready.
Resolve Processing Problems
| Problem | What to do |
|---|---|
| The upload is rejected | Check the supported formats and local upload limit. Enter a Document Title and review any validation message. |
| A PDF contains only scans or images | Use a text-based version. Collection uploads do not offer an OCR control. |
| A document shows Error | Check that the original file opens and contains readable text. Correct the file and upload it again. If it still fails, ask your administrator to check the collection's model configuration. |
| A document stays in processing | Refresh the list. If it remains unchanged, share the collection name, document title, and visible status with your administrator. |
| A web import fails | Check the URL and whether the content is accessible without your browser's signed-in session. Try a smaller crawl. |
| A storage import fails | Verify the endpoint, bucket, prefix, region, and credentials with your storage administrator. |
| The progress dialog cannot load status | Return to the collection and refresh its document list before starting another import. A status-loading error does not confirm that the import failed. |
| A completed document is missing from Playground suggestions | Check that you selected the correct project and have access to the collection, then search for the document or collection again. |
There is no reprocess button in the collection interface. After correcting a file, upload it again and check the new document's status. Remove the failed or outdated copy when you no longer need it and have permission to delete it.
Check the Result in the Playground
Select the completed document with @ and ask a question you can verify against the original:
Summarize the payment terms in this document and identify the supporting passages.
Review the answer against the source. For detailed source-selection steps, see Ask Questions About Documents.