Voyage-3-lite（ボヤージュスリーライト）

import voyageai

client = voyageai.Client(api_key="your-voyage-api-key")

# 単一テキストの埋め込み
result = client.embed(
    ["最新のGPUはどれが最も性能が高いですか？"],
    model="voyage-3-lite",
    input_type="query"  # "query" or "document"
)

vector = result.embeddings[0]  # 512次元リスト
print(f"次元数: {len(vector)}, トークン消費: {result.total_tokens}")

# バッチ埋め込み（ドキュメント側）
documents = [
    "RTX 4090は24GB GDDR6Xを搭載し、3DMarkスコアは30000超",
    "Radeon RX 7900 XTXは24GB GDDR6を搭載",
    "Arc B580は12GB GDDR6でコスパに優れる"
]

doc_embeddings = client.embed(
    documents,
    model="voyage-3-lite",
    input_type="document"
).embeddings

import anthropic
import voyageai
import numpy as np

voyage = voyageai.Client()
claude = anthropic.Anthropic()

def retrieve_and_generate(query: str, documents: list) -> str:
    # Step 1: 埋め込み生成
    query_emb = voyage.embed([query], model="voyage-3-lite",
                             input_type="query").embeddings[0]
    doc_embs = voyage.embed(documents, model="voyage-3-lite",
                            input_type="document").embeddings

    # Step 2: コサイン類似度で上位3件取得
    scores = [
        np.dot(query_emb, emb) / (np.linalg.norm(query_emb) * np.linalg.norm(emb))
        for emb in doc_embs
    ]
    top_indices = sorted(range(len(scores)), key=lambda i: scores[i], reverse=True)[:3]
    context = "\n\n".join(documents[i] for i in top_indices)

    # Step 3: Claudeで回答生成
    response = claude.messages.create(
        model="claude-opus-4-8",
        max_tokens=1024,
        messages=[{"role": "user", "content": f"Context:\n{context}\n\nQuestion: {query}"}]
    )
    return response.content[0].text