Notebooks
W
Weaviate Ollama Simple Tool Call

Ollama Simple Tool Call

vector-searchvector-databaseretrieval-augmented-generationOllamallm-frameworksfunction-callingweaviate-recipesintegrationsPythongenerative-aillm-agent-frameworks
alph-notebooks/weaviate-recipes / ollama-simple-tool-call.ipynb

Open In Colab

[17]
[35]
[19]
Search Result 1:
content:Let's explore how PQ works. ## Product Quantization
![ann](./img/Ann.png)

If you already know the details behind how Product Quantization works feel free to skip this section!

The main intuition behind Product Quantization is that it adds the concept of segments to the compression function. Basically, to compress the full vector, we will chop it up and operate on segments of it. For example, if we have a 128 dimensional vector we could chop it up into 32 segments, meaning each segment would contain 4 dimensions. Following this segmentation step we compress each segment independently.author:None
Search Result 2:
content:### Product Quantization
While async indexing helped with importing, [product quantization](/developers/weaviate/concepts/vector-quantization#product-quantization) (PQ) was important for alleviating the memory footprint. PQ is a technique used to compress the size of vector embeddings by representing the vectors as centroid ids. The centroid ids are calculated by clustering each segment. We enabled PQ to reduce the memory requirements without slowing down the ingestion. You can either [manually](/developers/weaviate/configuration/pq-compression#manually-configure-pq) enable PQ or have it [automatically triggered](/developers/weaviate/configuration/pq-compression#configure-autopq) based on the number of vectors.author:None
[20]
[33]
[34]
Calling function <function get_search_results at 0x114749a20>...

"Based on the search results provided, it appears that HNSW (Hierarchical Navigable Small World) is different from DiskANN in the following ways:\n\n1. **Implementation**: While both Vamana and HNSW are ANN algorithms used for vector indexing, Vamana is an implementation of DiskANN, whereas HNSW has its own highly optimized version implemented by Weaviate.\n2. **Performance**: Both Vamana (DiskANN) and the Weaviate-implemented HNSW achieve comparable in-memory results, indicating that both are effective for vector search tasks.\n3. **Challenges**: The implementation of DiskANN on disk-based vectors poses several challenges, including achieving good performance while maintaining excellent database UX.\n\nIn summary, while Vamana (DiskANN) and Weaviate's HNSW share similarities as ANN algorithms, they differ in their implementation, with HNSW having its own optimized version."