RAG

RAG (Smart Document Search + AI)

What It Does

RAG is like having a research assistant that actually reads your documents before answering questions. Instead of guessing, it searches through your knowledge base, finds relevant information, then uses AI to give you accurate, source-backed answers.

What Goes In, What Comes Out

Input

Name	Type	Description	Required	Default
llm	LLM Connection	Your AI model	Yes	-
vector_store	Vector Store	Your document database	Yes	-
query	Text	Question to ask	Yes	-
top_k	Number	How many documents to search	No	5
similarity_threshold	Number	How closely documents must match (0-1)	No	0.7

Output

Name	Type	Description
answer	Text	AI answer based on found documents
retrieved_documents	Array	Source documents used
confidence	Number	How confident the AI is (0-1)
sources	Array	Where the information came from

Real-World Examples

📚 Company Knowledge Base: Ask questions about policies, procedures, or documentation

• Input: “What’s our vacation policy?”
• Output: Accurate answer with policy references

🔍 Research Assistant: Get insights from large document collections

• Input: “What are the main findings about climate change?”
• Output: Summary with source citations

💬 Smart Customer Support: Answer questions using your help documentation

• Input: “How do I reset my password?”
• Output: Step-by-step instructions from your docs

How It Works

flowchart LR
    A[❓ Your Question] --> B[🔍 Search Documents]
    B --> C[📄 Find Relevant Info]
    C --> D[🤖 AI Analysis]
    D --> E[✅ Accurate Answer + Sources]

    style A fill:#e3f2fd
    style B fill:#fff3e0
    style C fill:#f3e5f5
    style D fill:#fff3e0
    style E fill:#e8f5e8

Why RAG is Better Than Regular AI:

• 🎯 More Accurate: Uses your actual documents, not AI’s training data
• 📚 Source Citations: Shows you exactly where answers come from
• 🚫 No Hallucinations: Can’t make up facts because it reads real documents first
• 🔄 Always Current: Uses your latest documents, not outdated training data

Quick Start Example

Goal: Create a smart FAQ system for your company docs

Setup:

1. Upload your documents to Local Knowledge
2. Connect RAG Node to search and answer
3. Ask questions like “What’s our return policy?”

Result: Get accurate answers with source references, just like having an expert who’s read all your documentation.

Configuration Tips

Essential Settings

• Top K: Start with 3-5 documents. More isn’t always better
• Similarity Threshold: 0.7 is good for most cases, 0.8+ for very specific matches
• Include Metadata: Turn on to see document titles and sources

Simple Setup Guide

For General Questions 💬

• Search 5 documents (good balance of speed and coverage)
• Set similarity to 0.7 (finds related content)
• Turn on metadata (shows document titles and sources)

For Precise Answers 🎯

• Search 3 documents (faster, more focused)
• Set similarity to 0.8 (very specific matches only)
• Use for technical or specific factual questions

For Research & Exploration 🔍

• Search 8 documents (comprehensive coverage)
• Set similarity to 0.6 (catches broader connections)
• Great for discovering related topics and concepts

Browser Compatibility

Works in all major browsers:

• ✅ Chrome: Full support with fast vector search
• ✅ Firefox: Full support
• ⚠️ Safari: Limited storage for large document collections
• ✅ Edge: Full support

Privacy & Security

• 🔒 Local Storage: Your documents stay in your browser
• 🔐 Encrypted: Document storage is encrypted for security
• 🚫 No External Sharing: Documents never leave your device
• ✅ Source Validation: Verifies document authenticity

Step-by-Step Workflow

1. Build Your Knowledge Base

Use Get All Text From Link + Local Knowledge to create your document collection

2. Ask Questions

Connect RAG Node and ask natural language questions

3. Get Smart Answers

Receive answers with source citations and confidence scores

4. Verify Sources

Check the retrieved documents to validate the information

Try It Yourself

Example 1: Company FAQ Bot

What you’ll build: Smart FAQ system that answers questions about your company

Workflow:

Get All Text From Link → Local Knowledge → RAG Node → Edit Fields

Setup:

1. Collect Documents: Use Get All Text to grab your FAQ pages, policies, etc.
2. Build Knowledge Base: Store everything in Local Knowledge
3. Configure RAG: Set similarity_threshold to 0.8 for precise answers
4. Ask Questions: “What’s our return policy?” → Get accurate, sourced answers

Result: Instant, accurate answers to company questions with source citations.

Example 2: Research Assistant

What you’ll build: AI that searches through research papers and gives sourced answers

Workflow:

Upload Documents → Local Knowledge → RAG Node → Download As File

Setup:

• Top K: 5 (to get comprehensive coverage)
• Similarity Threshold: 0.7 (to catch related concepts)
• Include Metadata: Yes (to see paper titles and dates)

Result: Ask “What are the main benefits of renewable energy?” and get a comprehensive answer with citations from your research collection.

🔍 Advanced Example: Multi-Language Knowledge Base

What you’ll build: Knowledge base that works across multiple languages

Setup:

• Use embedding models that support multiple languages
• Store documents in different languages in the same knowledge base
• RAG will find relevant documents regardless of language

Use case: International company with documentation in multiple languages.

Best Practices

✅ Do This

• Start with quality documents: Better source material = better answers
• Use descriptive document titles: Helps with source attribution
• Test similarity thresholds: 0.7 is good for most cases, adjust as needed
• Keep documents updated: Remove outdated information regularly

❌ Avoid This

• Storing too many irrelevant documents (creates noise)
• Setting similarity threshold too high (might miss relevant info)
• Asking questions outside your document scope
• Ignoring source citations (always verify important answers)

Troubleshooting

🎯 “No Relevant Documents Found”

Problem: RAG can’t find documents related to your question Solution: Lower similarity_threshold to 0.6 or add more documents to your knowledge base

🐌 Slow Search Results

Problem: RAG takes too long to find and process documents Solution: Reduce top_k to 3, or clean up your knowledge base to remove irrelevant documents

📄 Poor Answer Quality

Problem: Answers don’t make sense or miss important information Solution: Check if your documents actually contain the information you’re asking about

💾 “Storage Quota Exceeded”

Problem: Can’t add more documents to knowledge base Solution: Remove old/irrelevant documents or use document compression

Limitations to Know

• Document Quality Matters: RAG is only as good as the documents you feed it
• Storage Limits: Browser storage limits how many documents you can store
• Processing Time: Searching large document collections takes 2-5 seconds
• Question Scope: Can only answer questions about information in your documents

Related Nodes

🔄 Similar Nodes

• Q&A Node: Simpler question-answering without document search
• Basic LLM Chain: Basic AI processing without knowledge base

🔗 Works Great With

• Local Knowledge: Stores your documents for searching
• Recursive Character Text Splitter: Breaks documents into searchable chunks
• Ollama Embeddings: Creates searchable representations of your documents

🛠️ Required Setup

• Local Knowledge: Vector database for document storage
• Ollama Embeddings: For creating document embeddings
• Ollama or WbeLLM: AI model for generating answers

What’s Next?

🌱 New to Document Search?

Start with Local Knowledge to build your first document collection

🚀 Ready for More?

• Try Q&A Node for simpler question-answering
• Explore Vector Database Guide
• Check out real-world RAG examples

---

💡 Pro Tip: Start with a small, focused document collection (10-20 documents) to test your RAG setup, then gradually expand as you get comfortable with the results.