weaviate-search

README

Semantic Search with Zerfoo Embeddings

This example demonstrates using Zerfoo's embedding API to perform cosine-similarity semantic search over a small document corpus. The integrations/weaviate adapter wraps the /v1/embeddings endpoint and produces []float32 vectors suitable for insertion into Weaviate or any other vector database.

Prerequisites

• A compiled zerfoo binary (or go run ./cmd/zerfoo)
• An embedding model in GGUF format (e.g. nomic-embed-text)

Setup

1. Start the Zerfoo server with an embedding model:

zerfoo serve --model path/to/nomic-embed-text.gguf --port 8080

1. Run the search example:

go run ./examples/weaviate-search/ \
  --server http://localhost:8080 \
  --model nomic-embed-text \
  --query "How do I run ML inference in Go?" \
  --top-k 3

How It Works

The adapter (integrations/weaviate.Adapter) sends text to Zerfoo's /v1/embeddings endpoint and returns []float32 vectors. This example embeds a hardcoded corpus of 8 documents, then embeds the query and ranks documents by cosine similarity.

emb := weaviate.NewAdapter("http://localhost:8080", "nomic-embed-text")

// Embed documents
docVecs, err := emb.EmbedDocuments(ctx, documents)

// Embed a query
queryVec, err := emb.EmbedQuery(ctx, "How do I run ML inference in Go?")

In production, you would store docVecs in Weaviate using its Go client and use queryVec for near-vector searches.

Using with Weaviate

To insert vectors into a live Weaviate instance:

import "github.com/weaviate/weaviate-go-client/v4/weaviate"

client, _ := weaviate.NewClient(weaviate.Config{Host: "localhost:8080", Scheme: "http"})

// Use the adapter to get vectors, then batch-import into Weaviate
vecs, _ := emb.EmbedDocuments(ctx, documents)
for i, doc := range documents {
    client.Data().Creator().
        WithClassName("Document").
        WithProperties(map[string]any{"text": doc}).
        WithVector(vecs[i]).
        Do(ctx)
}

Flags

Flag	Default	Description
--server	http://localhost:8080	Zerfoo server URL
--model	nomic-embed-text	Embedding model name
--query	How do I run ML inference in Go?	Search query
--top-k	3	Number of results to return

Documentation

Overview

Command weaviate-search demonstrates using the Zerfoo Weaviate adapter to embed a corpus of documents and perform cosine-similarity semantic search without requiring a live Weaviate instance.

Start a Zerfoo server first:

zerfoo serve --model path/to/embed-model.gguf --port 8080

Then run this example:

go run ./examples/weaviate-search/ --server http://localhost:8080 --model nomic-embed-text