Workflows
Blocks
Database
Vector Database

Vector Store Block

What it does: Store and search text using AI-powered semantic understanding, finding relevant content based on meaning rather than exact keywords.

🔍

In simple terms: Think of a Vector Store as a smart librarian who understands what you're really looking for. Instead of searching for exact words, it finds content that means similar things. Ask "How do I return an item?" and it finds your return policy, even if it doesn't use the word "return."

When to Use This

Use a Vector Store when you need:

  • ✅ Semantic search (finding by meaning, not keywords)
  • ✅ Build a knowledge base for AI assistants
  • ✅ Power question-answering systems
  • ✅ Find similar documents or content
  • ✅ Enable RAG (Retrieval Augmented Generation) workflows

Example: Create a customer support bot that searches your help documentation and finds the most relevant articles to answer questions.

Key Features

  • Semantic Understanding: Find content by meaning, not exact matches
  • Similarity Scoring: See how relevant each result is
  • Fast Retrieval: Quick searches even with large document collections
  • AI Integration: Perfect for feeding context to AI models
  • Scalable Storage: Handle thousands of documents

How Vector Search Works (Simply Explained)

  1. Text becomes numbers: Your documents are converted into numerical representations (embeddings)
  2. Similar meanings, similar numbers: Content with similar meanings has similar numerical patterns
  3. Search by similarity: When you search, your query is converted the same way and compared
  4. Best matches returned: Documents with the closest numerical patterns are your results

Setup Guide

Step 1: Choose Your Vector Database

You can use the built-in Vector Store or connect to external providers like Pinecone.

Step 2: Add Your Content

Before searching, you need to populate your Vector Store with documents. You can add:

  • Text documents and articles
  • FAQ entries
  • Product descriptions
  • Knowledge base content
  • Any text you want to make searchable

Step 3: Configure Search Settings

SettingWhat It DoesRecommendation
QueryThe search textKeep it natural language
Number of ResultsHow many matches to return3-5 for most use cases
Similarity ThresholdMinimum relevance score0.7 for balanced results

Common Use Cases

AI Knowledge Base

Build a smart knowledge base that AI assistants can query to provide accurate, contextual answers based on your actual documentation.

Customer Support Search

Help customers find relevant help articles by understanding their questions, even when they use different words than your documentation.

Document Discovery

Find related documents within large collections - great for research, compliance, or finding similar contracts.

Content Recommendation

Suggest similar articles, products, or content based on what users are currently viewing.

RAG (Retrieval Augmented Generation)

Feed relevant context to AI models so they can answer questions accurately using your specific data.

Operations Available

Add Documents

Store new text content in your Vector Store, making it searchable.

What you provide:

  • The text content
  • Optional metadata (title, category, source)
  • Unique identifier for later reference

Search/Query

Find relevant content based on a natural language query.

What you provide:

  • Your search query
  • Number of results wanted
  • Optional filters

Delete Documents

Remove content from your Vector Store when it's no longer needed.

What You Get Back

After searching your Vector Store:

  • Matching Documents: The relevant text content found
  • Similarity Scores: How closely each result matches (0 to 1)
  • Metadata: Any additional info stored with each document
  • Document IDs: Unique identifiers for each result

Tips for Success

  1. Chunk your content - Break large documents into smaller, focused sections
  2. Add helpful metadata - Include source, date, and category for filtering
  3. Use natural language queries - Write searches like you're asking a person
  4. Review similarity scores - Low scores may indicate irrelevant results
  5. Update regularly - Keep your knowledge base fresh and current

Troubleshooting

ProblemLikely CauseSolution
No results foundEmpty store or very specific queryVerify content is indexed, broaden query
Irrelevant resultsQuery too vague or content mismatchMake query more specific or adjust threshold
Slow searchesVery large document collectionConsider pagination or filtering
Duplicate resultsSame content indexed multiple timesDe-duplicate before adding

Best Practices

  • Quality over quantity - Well-written, focused content gives better results
  • Consistent formatting - Structure similar documents the same way
  • Meaningful chunks - Each piece should make sense on its own
  • Regular maintenance - Remove outdated content, add new material
  • Test your searches - Try real user queries to verify quality
Indite Documentation v1.4.0
PrivacyTermsSupport