Unrag
DebuggingExperimental
Experimental Feature

Docs Panel

Document inspection and management for exploring your vector store.

The Docs panel lets you browse what's actually in your vector store. You can see which documents are stored, how they were chunked, what the chunks contain, and—when needed—delete content that shouldn't be there.

Setting it up

The Docs panel needs a bit more than the other panels to work. In addition to registering your engine, you also need to provide the store inspector: registerUnragDebug({ engine, storeInspector: engine.storeInspector }). The store inspector gives the debug server access to methods for listing and inspecting documents.

Without it, the panel shows a message explaining what's missing. Once it's configured, you're ready to explore.

Browsing your documents

The left side shows a list of all documents in your store. Each row displays the source ID, how many chunks the document has, and when it was created (if that information is available). Navigate with j and k, and use [ and ] to page through if you have more documents than fit on screen.

At the top, there's a store stats bar showing aggregate information: which adapter you're using, how many vectors are in the store, the embedding dimension, and row counts for the relevant tables. This gives you a quick sense of the overall state of your data.

Working with prefixes

If you have a lot of documents, the prefix filter helps you narrow down what you're looking at. Press f to edit the filter, type a prefix like docs: or user:123:, and press Enter. Now the list shows only documents whose source ID starts with that prefix.

This is particularly useful when you use a consistent naming convention for source IDs. Filter to a specific namespace to focus on just the content you care about.

Inspecting chunks

When you select a document, the right side updates to show its details. You'll see a token histogram—a visual distribution of how large the chunks are—which helps you understand whether your chunking is consistent or all over the place.

Below that is the chunk list, showing each chunk with its sequence number, approximate token count, and a preview of its content. Use Tab to switch focus between the documents list and the chunks list, then navigate chunks with j and k as usual.

The selected chunk's full content appears at the bottom, scrollable with PageUp and PageDown. This is where you verify that your chunks contain complete, coherent content—that sentences aren't cut off mid-thought and that related information stays together.

Deleting content

During development, you'll often need to remove documents—maybe you ingested test data, or a document was chunked incorrectly and you want to try again with different settings.

Press d on a selected document to delete it (the panel asks for confirmation first). This removes the document and all its chunks. If you only want to remove a specific chunk while keeping the rest, switch focus to the chunks list and press x on the chunk you want to delete.

Be careful with deletion—it's permanent. The debug panel doesn't have undo. If you're connected to a shared or production-like environment, think twice before confirming.

When content is missing

If your store adapter is configured with storeChunkContent: false, you'll see a warning that chunks have no stored content. In this mode, only the embeddings are persisted, not the original text. The chunk list will show metadata but empty content areas.

To use the Docs panel's full inspection capabilities, you'll need to enable content storage in your engine configuration.

On this page

RAG handbook banner image

Free comprehensive guide

Complete RAG Handbook

Learn RAG from first principles to production operations. Tackle decisions, tradeoffs and failure modes in production RAG operations

The RAG handbook covers retrieval augmented generation from foundational principles through production deployment, including quality-latency-cost tradeoffs and operational considerations. Click to access the complete handbook.