Performs fast k-NN search and clustering on dense vectors using FAISS index types.
npx skills add https://github.com/Quill-Agent/Quill-Agent --skill faiss-quill-agent
Please help me install this Agent Skill. Skill: faiss Source: https://github.com/Quill-Agent/Quill-Agent/tree/main/optional-skills/mlops/faiss Command: npx skills add https://github.com/Quill-Agent/Quill-Agent --skill faiss-quill-agent
This Skill provides a powerful and efficient tool for similarity search and clustering of dense vectors, supporting billions of vectors and GPU acceleration for optimal performance.
Install the FAISS Skill and create an index to perform a search:
pip install faiss-cpu
python
import faiss
index = faiss.IndexFlatL2(128) # 128 dimensions
index.add(np.random.random((1000, 128)).astype('float32'))
distances, indices = index.search(np.random.random((1, 128)).astype('float32'), 5)
High-intent search queries and answers about installing and using this skill.
Fast k-NN search on large-scale dense vectors is performed by creating a Flat, IVF, HNSW, or Product Quantization index, adding your dataset, and calling the search method to retrieve the nearest neighbors.
Similarity search for billions of vectors is handled using index types like IVF or Product Quantization to optimize memory and performance. This allows efficient clustering and retrieval without managing metadata.
GPU acceleration works for vector search and clustering tasks by utilizing the faiss-gpu library. This significantly increases processing speed for building indexes and querying dense vectors compared to CPU-only execution.
You need numpy to run large-scale vector retrieval operations because input vectors must be passed as contiguous numpy arrays. The library requires faiss-cpu or faiss-gpu and numpy to execute similarity searches.
Product Quantization should be used instead of a Flat index for similarity search when you need to compress large-scale dense vectors to fit into memory. Flat indexes provide exact results but do not scale to billions of vectors efficiently.
You can use this approach for similarity search applications without metadata, as it focuses purely on dense vector matching and clustering. It is ideal for scenarios requiring fast retrieval based solely on vector distances.
Awannaphasch2016Perform semantic vector search against AgentDB stores using embedding similarity.
PthahnixEnrich Scholar results with arXiv, S2, and Unpaywall metadata into a PaperMeta payload.
devMoezBuild and search FAISS vector indexes for nearest-neighbor retrieval.
FalkorDBCreate and query HNSW vector indexes in FalkorDB graph databases.