Prepare local retrieval embeddings with FastEmbed
Generate dense, sparse, image, and reranking embeddings locally before writing vectors into Qdrant or another retrieval stack for agent memory and RAG workflows.
npx skills add agentskillexchange/skills --skill prepare-local-retrieval-embeddings-with-fastembed
Use FastEmbed when an agent workflow needs a lightweight local embedding step before retrieval, memory, or RAG indexing. The operator installs FastEmbed, selects a supported dense, sparse, image, late-interaction, or reranker model, embeds the bounded document or image set locally, and sends vectors and payloads into Qdrant or another retrieval pipeline for later agent use. Invoke this instead of normal library use when the task is to prepare or refresh a retrieval index with inspectable local embedding generation. The boundary is retrieval-prep embedding and reranking; do not present FastEmbed as a generic Python embedding library.
What this skill actually does
Inputs and prerequisites: Python, FastEmbed, optional qdrant-client and Qdrant vector database.
Setup notes: Install with pip install fastembed or pip install fastembed-gpu for GPU support. For Qdrant ingestion, install qdrant-client[fastembed], create embeddings with FastEmbed models, then write vectors and payloads into the retrieval collection.
Source and verification boundary: use https://qdrant.github.io/fastembed/ 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.