User Guide¶

This guide walks you through every feature of Beyond Retrieval v2 from a user's perspective.

Getting Started¶

Opening the Application¶

• Local development: Navigate to http://localhost:3000
• Hosted deployment: Navigate to the domain provided by your administrator

Signing In¶

If authentication is enabled, you'll see a sign-in page. If bypassed (common in dev), you'll land directly on the Dashboard.

The Dashboard¶

The Dashboard shows a grid of all notebooks you have access to. From here you can create new notebooks, open existing ones, and access Global Settings from the sidebar.

Notebooks¶

Notebooks are the core organizational unit. Each notebook is an independent workspace with its own documents, conversations, search tools, and settings.

Creating a Notebook¶

1. Click Create Notebook on the Dashboard.
2. Fill in:

Choosing an Embedding Model¶

Documents¶

Uploading Files¶

Supported formats: PDF, DOCX, TXT, MD, CSV, XLSX, XLS, PPTX, HTML

1. Navigate to the Documents page inside your notebook.
2. Drag and drop files or click the upload area.
3. After uploading, click Ingest to start processing.

Ingestion Settings¶

Document Status¶

OneDrive Integration¶

If configured, click the OneDrive button on the Documents page to browse SharePoint folders and import files directly.

Chat (RAG Mode)¶

Starting a Conversation¶

1. Open a notebook → Chat page.
2. Click + New Chat.
3. Type your question and press Enter.

The system searches your documents, retrieves relevant chunks, and generates a cited answer with numbered references like [1], [2].

Citations¶

Click on citation numbers to see the source chunk text. Click the chunk to open the full document viewer with highlighted lines.

Personas¶

Languages¶

10 supported: English, German, Spanish, French, Italian, Portuguese, Dutch, Russian, Chinese, Japanese.

Language Mode:

• Auto-detect — mirrors the language of your question
• Manual — always responds in the selected language

Feedback¶

Thumbs up/down on each response. Used by the LLM Judge for quality tracking and caching decisions.

Search Playground¶

Test and compare retrieval strategies without using chat.

Available Strategies¶

Compare Mode¶

Run up to 4 strategies on the same query side-by-side to see which retrieves the most relevant content.

AI Enhancer¶

Enriches document chunks with AI-generated context for better search accuracy.

Workflow¶

1. Not Enhanced — Select files to enhance
2. Enhancing — AI processes chunks in parallel (progress updates every 4s)
3. Published — Enhanced chunks are live in the vector store

Click Enhance → wait for all chunks to show "success" → click Publish.

Intelligence Settings¶

Per-notebook AI configuration:

System Monitor¶

Health overview and maintenance tools:

• Health Score (0–100) — based on duplicates, orphans, and enhanced chunks
• Cleanup Tools — Remove Duplicates, Remove Orphans
• Document Statistics — Total, Indexed, Errors, Pending

Global Settings (Admin Only)¶

• API Keys — Configure OpenRouter, OpenAI, Mistral, Cohere keys
• Database Type — Switch between Cloud and Local Supabase
• Storage Provider — Supabase, S3, Local, or None

Sharing¶

1. Open notebook → sharing settings
2. Generate Invite Link with access level (Admin or Chat Only)
3. Share the link — recipients sign in and accept to gain access

Admins can view, revoke access, and deactivate invite codes.

Troubleshooting¶