Skill Detail

Build local vector retrieval indexes with Faiss

Use Faiss to create and query local vector indexes for agent retrieval and RAG workflows before adding heavier managed vector infrastructure.

Data Extraction & TransformationMulti-Framework
Data Extraction & Transformation Multi-Framework Security Reviewed
โญ 40.2k GitHub stars
INSTALL WITH ANY AGENT
npx skills add agentskillexchange/skills --skill build-local-vector-retrieval-indexes-with-faiss Copy
Works best when you want a reusable capability, not another fragile one-off prompt.
At a glance
Tools required
Faiss, embeddings, Python or native runtime bindings
Install & setup
Install the appropriate Faiss package or build from source, prepare embedding vectors, create a Faiss index, add vectors with IDs, and query the index from the agent retrieval workflow.
Author
Meta AI / Facebook Research
Publisher
Open Source
Last updated
Jun 9, 2026
Quick brief

Use this when an agent or operator needs to build a local retrieval index from embeddings, tune nearest-neighbor search behavior, and hand back ranked candidates for RAG, semantic search, or offline evaluation. The workflow is bounded to indexing and querying vectors: prepare embeddings, choose an index type, add vectors with stable IDs, run similarity search, and return evidence for downstream agent steps. It is useful when a workflow needs fast local experiments, offline retrieval checks, or reproducible comparison against heavier vector database choices. This is not a generic Faiss library card; it is for local vector retrieval operations where the agent needs repeatable search setup and inspection instead of relying on a hosted vector database.

How it works

What this skill actually does

Inputs and prerequisites: Faiss, embeddings, Python or native runtime bindings.

Setup notes: Install the appropriate Faiss package or build from source, prepare embedding vectors, create a Faiss index, add vectors with IDs, and query the index from the agent retrieval workflow.

Source and verification boundary: use https://github.com/facebookresearch/faiss as the canonical reference before running the workflow; keep commands, API calls, CLI usage, and generated outputs reviewable against that upstream source.

Framework fit: publish this as a Multi-Framework workflow only when the operator can invoke the documented toolchain directly, rather than treating the upstream project as a generic product listing.