Documents

Upload, browse, and manage documents in your knowledge base.

The Documents tab is the main document management interface. It has six sub-tabs for different aspects of document handling.

Library

Browse all documents in your knowledge base:

• File name, size, and upload date

• Processing status (pending, processing, completed, failed)

• Chunk count — how many searchable segments were created

• Actions — view, re-process, delete

File types supported

Type	Extensions
Text	.md, .txt, .csv, .json
Office	.docx, .xlsx, .pptx
PDF	.pdf (including scanned documents via OCR)
Code	.py, .js, .ts, .go, .java, and more
Web	.html

Upload

Bulk upload interface:

1. Click Upload Documents or drag files onto the page

2. Files are queued for processing

3. The ingestion pipeline runs automatically:

   • Parse — extract text from the file format

   • Chunk — split into semantic segments

   • Embed — generate vector embeddings

   • Index — store in the vector database

Large files are processed asynchronously. Check the Processing sub-tab for status updates.

Processing

Monitor the ingestion pipeline:

• Files in queue

• Currently processing (with progress)

• Recently completed

• Failed (with error details)

Search

Test semantic search across your knowledge base:

1. Type a natural language query

2. See ranked results with relevance scores

3. Click a result to view the source document and chunk

This is exactly what your agents see when they search your knowledge base during chat.

RAG

Configure retrieval-augmented generation settings:

• Top K — number of chunks retrieved per query

• Similarity threshold — minimum relevance score

• Reranking — enable/disable result reranking

• Context window — how much surrounding text to include

Multimodal

Documents with rich media:

• Images with extracted text (OCR)

• PDFs with embedded graphics

• Mixed-content documents

The embedding model used is configurable in Settings. Default is qwen/qwen3-embedding-8b with 2048 dimensions.