faiss
by facebook
GPU-accelerated billion-scale vector similarity search and clustering
facebook/faissmixpeek://vector_index@v1/facebook_faiss_v1Overview
FAISS is Meta's library for efficient similarity search and clustering of dense vectors. It supports multiple index types (IVF, PQ, HNSW, flat), product quantization for compression, and optimized GPU kernels that handle billion-scale datasets.
On Mixpeek, FAISS powers the vector search infrastructure behind retriever stages, enabling sub-millisecond approximate nearest neighbor queries over large embedding collections.
Architecture
C++ library with Python bindings. Supports flat (exact), IVF (inverted file), PQ (product quantization), HNSW (graph-based), and composite indexes. GPU batched search with CUDA kernels for billion-scale workloads.
Mixpeek SDK Integration
# FAISS is used internally by Mixpeek's vector store.
# You don't call it directly — it powers the search infrastructure.
import { Mixpeek } from "mixpeek";
const mx = new Mixpeek({ apiKey: "API_KEY" });
// Ingest content — FAISS indexes the embeddings automatically
await mx.collections.ingest({
collection_id: "my-collection",
source: { url: "https://example.com/data.mp4" },
feature_extractors: [{
name: "image_embedding",
version: "v1"
}]
});Capabilities
- Billion-scale approximate nearest neighbor search
- GPU-accelerated with CUDA kernels
- Product quantization for 10-100x memory compression
- Multiple index types: IVF, PQ, HNSW, flat
- Clustering with k-means at scale
Use Cases on Mixpeek
Specification
Build a pipeline with faiss
Add this model to a processing pipeline alongside other extractors. Combine with retrieval stages for end-to-end search.
Open Pipeline Builder