For semantic search, vectorization turns text or code into numerical embeddings that can be compared by distance. Encode stored items and queries with the same embedding model and dimensional space; mismatches make comparisons unreliable. section 4.4.2 section 10.2.2
A practical retrieval workflow:
- Encode documents or code, then store the vectors alongside useful metadata in a vector database. section 10.2.2
- Encode a query with the same model and retrieve nearby vectors. In RAG, selected matches can become examples or context in a prompt; filter them to fit the available context window. section 4.4.2 section 10.2.3
Finding related coding instructions
Encode Python coding instructions with a code-search embedding model, index them in Milvus, then encode a new request and retrieve nearby instructions. The matches can provide examples for a coding assistant’s prompt. section 10.2.2 section 10.2.3
Listing 4.11 Example embeddings
import numpy as np
from sentence_transformers import SentenceTransformer
from datasets import load_dataset
model_ckpt = "sentence-transformers/all-MiniLM-L6-v2" #1
model = SentenceTransformer(model_ckpt)
embs_train = load_dataset("tweet_eval", "emoji", split="train[:1000]")
embs_test = load_dataset("tweet_eval", "emoji", split="test[:100]")
def embed_text(example): #2
embedding = model.encode(example["text"])
return {"embedding": np.array(embedding, dtype=np.float32)}
print(f"Train 1: {embs_train[0]}")
embs_train = embs_train.map(embed_text, batched=False)
embs_test = embs_test.map(embed_text, batched=False)
embs_train.add_faiss_index("embedding") #3
#
idx, knn = 1, 3 # Select the first query and 3 nearest neighbors #4
query = np.array(embs_test[idx]["embedding"], dtype=np.float32)
scores, samples = embs_train.get_nearest_examples("embedding", query, k=knn)
print(f"QUERY LABEL: {embs_test[idx]['label']}") #5
print(f"QUERY TEXT: {embs_test[idx]['text'][:200]} [...]\n")
print("=" * 50)
print("Retrieved Documents:")
for score, label, text in zip(scores, samples["label"], samples["text"]):
print("=" * 50)
print(f"TEXT:\n{text[:200]} [...]")
print(f"SCORE: {score:.2f}")
print(f"LABEL: {label}")
See a Python example that encodes tweet data with SentenceTransformer and searches for nearest neighbors using FAISS. section 4.4.2
