Inspecting Documents

Once a document has been processed, you can inspect it to see exactly how it was chunked, what content was extracted, and what metadata is available.

Viewing Document Details

From the Knowledge Base documents list, click on any document to open its detail view. Here you'll see:

• File name and upload timestamp
• Processing status (completed, failed, or in-progress)
• Total chunks extracted from the document
• Chunk list — a scrollable list of every chunk with its content

Chunk Information

Each chunk displays:

• Chunk index — sequential position in the document
• Content — the searchable text (or AI summary for tables/images)
• Original content — the raw extracted text before summarisation
• Element type — text, table, or image
• Page number — which page the chunk was extracted from (for PDFs)

Why Inspect Documents?

Inspecting is useful for:

• Quality assurance — verify that the partitioner correctly extracted your content
• Debugging — if the AI gives unexpected answers, check if the relevant content was chunked properly
• Understanding citations — see exactly what chunk the AI referenced in its response
• Tuning settings — if chunks are too large or too small, adjust the chunking parameters (in future versions)

Deleting Documents

To remove a document from your knowledge base, click the delete button on the document. This will:

1. Remove the document record from the database
2. Delete all associated chunks and embeddings
3. Remove the file from S3 storage

The AI will no longer reference this document in future conversations.