{"ok":true,"province":"vectors","name":"ベクトルDB","name_en":"Vector Database Decision","level":1,"criteria":[{"id":"v_cost","label":"コスト","label_en":"Cost","direction":"lower","description":null,"unit":"idx","kind":"editorial"},{"id":"v_selfhost","label":"自己ホスト","label_en":"Self-host","direction":"higher","description":null,"unit":null,"kind":"editorial"},{"id":"v_scale","label":"スケール","label_en":"Scale","direction":"higher","description":null,"unit":null,"kind":"editorial"},{"id":"v_ease","label":"手軽さ","label_en":"Ease","direction":"higher","description":null,"unit":null,"kind":"editorial"}],"items":[{"id":"vd_pinecone","name":"Pinecone","name_en":"Pinecone","provider":"Pinecone","jurisdiction":"US","note":"最易マネージド","note_en":"Easiest managed option","url":"https://www.pinecone.io/","shop_url":null,"metrics":{"v_cost":60,"v_selfhost":0,"v_scale":90,"v_ease":95},"facts":[],"sources":{"v_cost":"editorial: 自己ホストの選択肢が無くマネージドの利用料が必ずかかる型で、費用の値は候補8つで最も大きいため高め（2026-05初評）","v_selfhost":"editorial: マネージド専用のサービスで自前の基盤に置く形が用意されておらず、候補で唯一の最下位のため低め（2026-05初評）","v_scale":"editorial: 容量の拡張を運営側が肩代わりするマネージド型で、利用者は規模を意識せずに済み上位のため高め（2026-05初評）","v_ease":"editorial: サーバーの用意も運用も要らないフルマネージドで、始めやすさは候補8つで最上位のため高め（2026-05初評）"}},{"id":"vd_weaviate","name":"Weaviate","name_en":"Weaviate","provider":"Weaviate","jurisdiction":"EU","note":"ハイブリッド検索","note_en":"Hybrid search","url":"https://weaviate.io/","shop_url":null,"metrics":{"v_cost":45,"v_selfhost":90,"v_scale":75,"v_ease":70},"facts":[],"sources":{"v_cost":"editorial: OSSだがハイブリッド検索の運用はクラウド版に寄りやすく、費用の値は候補で二番目に大きいため中位（2026-05初評）","v_selfhost":"editorial: OSSでコンテナの配布があり自前の環境にそのまま置け、自己ホストは候補の中ほどの同点のため高め（2026-05初評）","v_scale":"editorial: 水平分割に対応するが大規模時は資源の消費が重くなりやすく、候補の中位下寄りのためやや高め（2026-05初評）","v_ease":"editorial: スキーマやモジュールの設計など最初に覚えることが多めで、手軽さは候補の中ほどのため中位の上（2026-05初評）"}},{"id":"vd_qdrant","name":"Qdrant","name_en":"Qdrant","provider":"Qdrant","jurisdiction":"EU","note":"無料枠最良・自己ホスト易","note_en":"Best free tier; easy to self-host","url":"https://qdrant.tech/","shop_url":null,"metrics":{"v_cost":30,"v_selfhost":95,"v_scale":80,"v_ease":80},"facts":[],"sources":{"v_cost":"editorial: 無料枠が手厚く軽量な自己ホストも選べる型で、費用の値は候補の中ほどより小さい側のためやや低め（2026-05初評）","v_selfhost":"editorial: OSSで単一コンテナから自前で動かせ構成も軽く、自己ホストのしやすさは候補上位の同点のため高め（2026-05初評）","v_scale":"editorial: 分散構成とシャーディングに対応するが超大規模の専用設計には譲り、候補の中ほどのため高め（2026-05初評）","v_ease":"editorial: 単一のコンテナで立ち上がり無料枠のクラウドも用意されていて、手軽さは候補上位のため高め（2026-05初評）"}},{"id":"vd_milvus","name":"Milvus","name_en":"Milvus","provider":"Zilliz","jurisdiction":"Global","note":"超大規模OSS","note_en":"OSS for very large scale","url":"https://milvus.io/","shop_url":null,"metrics":{"v_cost":15,"v_selfhost":95,"v_scale":95,"v_ease":50},"facts":[],"sources":{"v_cost":"editorial: OSSを自前で動かせば利用料はかからず、費用の値は候補8つの中で小さい側に入るため低め（2026-05初評）","v_selfhost":"editorial: OSSとして公開され自前の基盤に置け、部品は多いが構成の自由度が高く候補上位の同点のため高め（2026-05初評）","v_scale":"editorial: 保存と計算を分けた分散設計で超大規模を狙うOSSで、スケールはVespaに次ぐ上位のため高め（2026-05初評）","v_ease":"editorial: 複数の構成部品を組み合わせて動かす分散設計で導入が重く、手軽さは候補下位のため中位（2026-05初評）"}},{"id":"vd_chroma","name":"Chroma","name_en":"Chroma","provider":"Chroma","jurisdiction":"US","note":"小規模に最易OSS","note_en":"Easiest OSS for small-scale use","url":"https://www.trychroma.com/","shop_url":null,"metrics":{"v_cost":10,"v_selfhost":90,"v_scale":40,"v_ease":90},"facts":[],"sources":{"v_cost":"editorial: 小規模なら手元やアプリ内で動くOSSで追加の費用がほぼ要らず、費用の値は候補で最も小さいため低め（2026-05初評）","v_selfhost":"editorial: OSSで手元の端末から自前サーバーまで置き場所を選ばず、自己ホストは中ほどの同点のため高め（2026-05初評）","v_scale":"editorial: 小規模や試作を主眼にした軽量な作りで大規模な分散には向かず、スケールは候補で最下位のため低め（2026-05初評）","v_ease":"editorial: ライブラリを入れるだけで手元で動く小規模向けの作りで、手軽さはPineconeに次ぐ上位のため高め（2026-05初評）"}},{"id":"vd_pgvector","name":"pgvector","name_en":"pgvector","provider":"PostgreSQL","jurisdiction":"Global","note":"Postgres同居","note_en":"Lives inside Postgres","url":"https://github.com/pgvector/pgvector","shop_url":null,"metrics":{"v_cost":10,"v_selfhost":100,"v_scale":55,"v_ease":75},"facts":[],"sources":{"v_cost":"editorial: 既存のPostgresに同居する拡張で専用基盤を足さずに済み、費用の値は最も小さい同点のため低め（2026-05初評）","v_selfhost":"editorial: Postgresの拡張として既存のDBサーバーにそのまま載せられ、自己ホストは候補で最上位のため高め（2026-05初評）","v_scale":"editorial: 単一のPostgresの性能に縛られ、規模が大きくなると索引と資源が重くなり候補下位のため中位（2026-05初評）","v_ease":"editorial: 使い慣れたPostgresに拡張を足すだけで済むがDBの運用知識は要り、候補の中ほどのためやや高め（2026-05初評）"}},{"id":"vd_vespa","name":"Vespa","name_en":"Vespa","provider":"Vespa","jurisdiction":"US","note":"10億+規模","note_en":"1B+ scale","url":"https://vespa.ai/","shop_url":null,"metrics":{"v_cost":20,"v_selfhost":90,"v_scale":98,"v_ease":35},"facts":[],"sources":{"v_cost":"editorial: OSSで利用料は無いが大規模クラスタの運用が前提になる分だけ残り、費用の値は小さい側のため低め（2026-05初評）","v_selfhost":"editorial: OSSとして全機能を自前のクラスタで動かせる設計で、自己ホストは候補の中ほどの同点のため高め（2026-05初評）","v_scale":"editorial: 超大規模の検索と順位づけを分散処理する前提で設計され、スケールは候補8つで最上位のため高め（2026-05初評）","v_ease":"editorial: 独自の設定の書き方と配備手順を覚える必要があり学習の山が高く、手軽さは候補で最下位のため低め（2026-05初評）"}},{"id":"vd_opensearch","name":"OpenSearch","name_en":"OpenSearch","provider":"AWS","jurisdiction":"US","note":"検索基盤兼用","note_en":"Doubles as a search platform","url":"https://opensearch.org/","shop_url":null,"metrics":{"v_cost":35,"v_selfhost":90,"v_scale":80,"v_ease":55},"facts":[],"sources":{"v_cost":"editorial: 検索基盤と兼用できる反面クラスタの常時稼働が前提になり、費用の値は候補の中ほどのためやや低め（2026-05初評）","v_selfhost":"editorial: OSSの検索エンジンで自前のサーバーにもクラウドにも置け、自己ホストは中ほどの同点のため高め（2026-05初評）","v_scale":"editorial: 検索基盤として培った分散クラスタの仕組みをそのまま使え、Qdrantと同点の中ほどのため高め（2026-05初評）","v_ease":"editorial: 検索エンジンのクラスタ設定と索引設計が必要で導入の手数が多く、手軽さは候補下位のため中位（2026-05初評）"}}],"profiles":[{"id":"v_managed","name":"マネージド重視","name_en":"Managed","weights":{"v_ease":0.4000000000000001,"v_scale":0.30000000000000004,"v_cost":0.20000000000000004,"v_selfhost":0.10000000000000002}},{"id":"v_oss","name":"自己ホスト/OSS","name_en":"Open / self-host","weights":{"v_selfhost":0.4000000000000001,"v_cost":0.30000000000000004,"v_scale":0.20000000000000004,"v_ease":0.10000000000000002}},{"id":"v_scale","name":"大規模","name_en":"At scale","weights":{"v_scale":0.5000000000000001,"v_ease":0.20000000000000004,"v_cost":0.20000000000000004,"v_selfhost":0.10000000000000002}},{"id":"v_bal","name":"バランス","name_en":"Balanced","weights":{"v_cost":0.25,"v_selfhost":0.25,"v_scale":0.25,"v_ease":0.25}}],"articles":[{"slug":"what-is-rag","title":"RAGって結局なに？ 自分の資料でAIに答えさせる仕組みを、やさしく","dek":"AIは社内資料も最新情報も知らない。RAGは“答える前に調べさせる”ことで、出典つきの答えを返す仕組み。図でやさしく。"}],"terms":[],"campaigns":[]}