Migrating to Attanix: A Guide for Vector Store Users

Vector stores are a fundamental part of many modern AI applications, providing efficient similarity search capabilities. This guide shows how to enhance your vector store applications with Attanix's memory capabilities while maintaining compatibility with your existing code.

Why Add Attanix to Your Stack?

Attanix complements existing vector stores with additional capabilities:

1. Context-Aware Memory: Enhance similarity search with contextual understanding
2. Rich Metadata: Maintain detailed information about stored items
3. Adaptive Retrieval: Learn from usage patterns over time
4. Easy Integration: Compatible with existing vector store code

Integration Guide

Here's how to integrate Attanix with your existing stack:

from attanix import MemorySystem
from attanix.compat import VectorStoreAdapter

# Initialize Attanix with vector store compatibility
memory = MemorySystem()
store_adapter = VectorStoreAdapter(memory)

# Create an index
index = store_adapter.Index("your-index")

# Your existing code continues to work
index.upsert(vectors=[...])
results = index.query(vector=[...])

Compatibility Layer Features

1. Vector Operations

# Upsert vectors (same as Pinecone)
await index.upsert(
    vectors=[
        {"id": "1", "values": [0.1, 0.2, 0.3], "metadata": {"text": "example"}},
        {"id": "2", "values": [0.4, 0.5, 0.6], "metadata": {"text": "another"}}
    ]
)

# Query vectors (same as Pinecone)
results = await index.query(
    vector=[0.1, 0.2, 0.3],
    top_k=5,
    include_metadata=True
)

2. Metadata Handling

# Store with metadata (enhanced)
await index.upsert(
    vectors=[{
        "id": "doc1",
        "values": [...],
        "metadata": {
            "text": "content",
            "context": {
                "source": "web",
                "timestamp": "2024-04-18",
                "importance": 0.8
            }
        }
    }]
)

3. Advanced Queries

# Query with context (Attanix enhancement)
results = await index.query(
    vector=[...],
    context={
        "timeframe": "last week",
        "importance_threshold": 0.7
    }
)

Migration Strategies

1. Phased Integration

class HybridIndex:
    def __init__(self, vector_store, attanix_index):
        self.vector_store = vector_store
        self.attanix = attanix_index
    
    async def upsert(self, vectors):
        # Write to both during migration
        await self.vector_store.upsert(vectors)
        await self.attanix.upsert(vectors)
    
    async def query(self, vector):
        # Combine results during testing
        store_results = await self.vector_store.query(vector)
        attanix_results = await self.attanix.query(vector)
        return self.merge_results(store_results, attanix_results)

2. Data Migration

async def migrate_data(vector_store, attanix_index):
    # Fetch all vectors from Pinecone
    vectors = await vector_store.fetch_all()
    
    # Transform and store in Attanix
    for batch in chunk_vectors(vectors, 100):
        await attanix_index.upsert(batch)

3. Feature Migration

async def migrate_features():
    # Map Pinecone features to Attanix
    feature_map = {
        "namespace": "context",
        "filter": "metadata",
        "include_values": "include_vectors"
    }
    
    # Update queries to use new features
    updated_queries = migrate_queries(existing_queries, feature_map)

Best Practices

1. Testing

   • Compare query results
   • Verify performance metrics
   • Test edge cases

2. Monitoring

   • Track migration progress
   • Monitor system performance
   • Alert on discrepancies

3. Optimization

   • Tune Attanix parameters
   • Optimize batch sizes
   • Configure appropriate thresholds

Real-World Examples

Here are some successful migrations:

• Search Applications: Enhanced relevance with context
• Recommendation Systems: Better personalization
• Content Moderation: Improved accuracy with salience

Next Steps

Ready to migrate from Pinecone to Attanix? Check out our migration guide or try our compatibility layer.