RAG System

Question-Based Document Search with LLM

Backend Connected

Upload Documents

Click to upload or drag and drop

Files or entire folders (PDF, DOCX, TXT, HTML, CSV, XLSX, ZIP)

🔍 Nur Chunks anzeigen (nicht speichern)

Uploading...

Settings

Collection

Manage document collections

Options for processing uploaded documents

Docling Processing

Advanced document structure extraction for PDFs and Office files

Question Generation

Generate semantic questions per chunk using LLM (improves matching)

Semantic Chunking

Split documents by semantic similarity (+19% NDCG)

Real-time improvements applied during queries

Hybrid Search

Combine BM25 keyword + vector search (+31% NDCG)

Query Expansion

Generate query variations for better recall (+6% NDCG)

Reranking

Cross-Encoder reranking for best results (+47% NDCG)

Similarity Threshold

0.00 0.70 1.00

Filter out low-similarity chunks

System Prompts (RAG)

With documents No documents

Chunk Size (characters)

Size of each text chunk.

Chunk Overlap (characters)

Overlap between chunks

📥 Import Pre-chunked JSON

Apply processing

Use Docling Generate questions

Active Embedding Model

Loading...

Queries use this model for the current collection

Use Documents (RAG)

Show Thinking Steps

Top K Results

1 10 50

History Turns (Q/A pairs)

0 disables history. Stored only in this browser.

Documents

No documents uploaded

Total Documents: 0

Total Chunks: 0

Start a Conversation

Upload documents and ask questions to get started

Create New Collection

Collection Name *

Description (optional)

Embedding Model *

Documents in this collection will use this embedding model

Delete Collection

Are you sure you want to delete the collection ?

⚠️ This action cannot be undone. All documents and chunks in this collection will be permanently deleted.

📄 Chunks Preview

⚙️ Active Settings: