# Qdrant MCP Server Examples
Practical examples demonstrating how to use the Qdrant MCP Server for various use cases.
## Quick Start
Before running these examples, ensure you have:
1. **Qdrant running**: `docker compose up -d`
2. **MCP Server configured**: See main [README](../README.md) for setup
3. **OpenAI API key**: Set in your environment or MCP configuration
## Available Examples
### 🎯 [Basic Usage](./basic/)
**Start here if you're new to the MCP server.**
Learn fundamental operations:
- Creating collections
- Adding documents
- Performing semantic searches
- Managing resources
**Time:** 5-10 minutes
**Difficulty:** Beginner
---
### ⚡ [Rate Limiting](./rate-limiting/)
**Understand automatic rate limit handling for batch operations.**
Topics covered:
- Configuring rate limits for your OpenAI tier
- Batch document processing
- Exponential backoff retry behavior
- Monitoring and troubleshooting
**Use cases:**
- High-volume document ingestion
- Free tier optimization
- Production reliability
- Batch operations
**Time:** 10-15 minutes
**Difficulty:** Beginner to Intermediate
---
### 📚 [Knowledge Base](./knowledge-base/)
**Build a searchable documentation system with metadata.**
Topics covered:
- Structuring documents with rich metadata
- Organizing content by team, topic, and difficulty
- Filtering searches by categories
- Scaling and maintenance strategies
**Use cases:**
- Company documentation
- Help centers
- Internal wikis
- Educational content
**Time:** 15-20 minutes
**Difficulty:** Intermediate
---
### 🔍 [Advanced Filtering](./filters/)
**Master complex search filters with boolean logic.**
Learn to:
- Combine multiple filter conditions (AND, OR, NOT)
- Filter by categories, prices, ratings, and availability
- Use range filters for numeric values
- Build sophisticated e-commerce searches
**Use cases:**
- Product catalogs
- Inventory management
- Content filtering
- Access control
**Time:** 20-30 minutes
**Difficulty:** Intermediate to Advanced
---
## Example Structure
Each example includes:
- **README.md** - Step-by-step tutorial with commands
- **Sample data** - Documents with metadata to copy/paste
- **Search examples** - Real queries you can try
- **Best practices** - Tips and recommendations
## How to Use These Examples
### Option 1: Interactive with Claude Code
Copy commands directly into Claude Code:
```
Create a collection named "test"
Add documents...
Search for...
```
### Option 2: Understand the Concepts
Read through examples to understand:
- Metadata design patterns
- Filter syntax and structure
- Search result ranking
- Collection organization
### Option 3: Adapt for Your Use Case
Use examples as templates:
1. Start with the closest matching example
2. Modify metadata structure for your data
3. Adjust search queries for your needs
4. Scale up with your own documents
## Learning Path
```
┌─────────────┐
│ Basic │ Start here - core operations
└──────┬──────┘
│
├──────────┐
│ ▼
│ ┌─────────────┐
│ │ Rate │ Learn automatic rate limit handling
│ │ Limiting │ (especially for batch operations)
│ └─────────────┘
│
▼
┌─────────────┐
│ Knowledge │ Add structure with metadata
│ Base │
└──────┬──────┘
│
▼
┌─────────────┐
│ Advanced │ Master complex filtering
│ Filtering │
└─────────────┘
```
## Common Patterns
### Pattern 1: Content Organization
```
metadata: {
"category": "type",
"topic": "subject",
"author": "creator",
"date": "2024-01-15"
}
```
Good for: Blogs, documentation, articles
### Pattern 2: E-commerce
```
metadata: {
"category": "electronics",
"price": 99.99,
"rating": 4.5,
"in_stock": true,
"brand": "BrandName"
}
```
Good for: Product catalogs, inventory systems
### Pattern 3: Access Control
```
metadata: {
"visibility": "internal",
"department": "engineering",
"sensitivity": "confidential"
}
```
Good for: Enterprise knowledge bases, secure documents
### Pattern 4: Temporal Data
```
metadata: {
"created_at": "2024-01-15",
"updated_at": "2024-03-20",
"status": "published",
"version": "2.1"
}
```
Good for: Versioned content, time-series data
## Tips for All Examples
### 1. Start Small
Test with 5-10 documents before scaling to thousands.
### 2. Design Metadata First
Plan your metadata structure before adding documents:
- What fields do you need?
- What will you filter by?
- What types (string, number, boolean)?
### 3. Use Consistent IDs
Choose an ID scheme:
- Sequential: `1`, `2`, `3`
- Prefixed: `doc-001`, `doc-002`
- Semantic: `eng-api-auth`, `sales-pricing`
- UUIDs: Generated automatically
### 4. Test Searches
Try various queries to validate:
- Semantic matching works correctly
- Filters return expected results
- Result ranking makes sense
### 5. Clean Up
Delete test collections when done:
```
Delete collection "collection-name"
```
## Troubleshooting
### Collection Already Exists
```
Delete collection "name"
Create a collection named "name"
```
### No Search Results
- Check collection has documents: `Get info about "collection-name"`
- Verify filter syntax matches metadata exactly
- Try search without filters first
### Unexpected Results
- Check document metadata structure
- Validate filter conditions
- Review semantic similarity scores
### Error Messages
- **"Collection not found"**: Create the collection first
- **"Bad Request"**: Check filter JSON syntax
- **"OpenAI API error"**: Verify API key and credits
## Next Steps
After completing these examples:
1. **Read the main [README](../README.md)** - Full tool documentation
2. **Explore your own data** - Apply concepts to real use cases
3. **Check out advanced features** - Quantization, hybrid search (future)
4. **Join the community** - Contribute examples and improvements
## Contributing
Have a great example to share? Please open a PR!
We're especially interested in:
- Domain-specific use cases (legal, medical, finance)
- Integration examples (RAG pipelines, chatbots)
- Performance optimization patterns
- Novel metadata structures
## Additional Resources
- [Qdrant Documentation](https://qdrant.tech/documentation/)
- [OpenAI Embeddings Guide](https://platform.openai.com/docs/guides/embeddings)
- [Model Context Protocol](https://modelcontextprotocol.io/)
- [Main Project README](../README.md)
