{
  "format": "agent-article-v0",
  "article_id": "EA-067",
  "canonical_url": "/tools/ea-067-knowledge-freshness-engineering",
  "page_type": "reference",
  "claim_ids": [],
  "evidence_ids": [
    "EVD-EA-067-URL-E0E9622033AF",
    "EVD-EA-067-URL-E8CFEB3458D9",
    "EVD-EA-067-URL-D320C481D01C",
    "EVD-EA-067-URL-7C7C7DBBC222",
    "EVD-EA-067-URL-883737C6EFB0",
    "EVD-EA-067-URL-7BBD2F9AC09C",
    "EVD-EA-067-URL-5C9DF414C214",
    "EA-067-R11-RESEARCH",
    "EA-067-R11-WEB-READBACK"
  ],
  "unknown_ids": [
    "EA-067-UNKNOWN-01",
    "EA-067-UNKNOWN-CURRENTNESS"
  ],
  "operational_object_ids": [],
  "provider": [],
  "models": [],
  "plans": [],
  "surfaces": [],
  "regions": [],
  "current_status": "UNKNOWN",
  "observation_status": "SPEC_ONLY",
  "freshness": "UNKNOWN",
  "last_verified": null,
  "supersedes": [],
  "superseded_by": [],
  "provenance": [
    "COMM-063-EA0067",
    "SRC-EA-067-URL-E0E9622033AF",
    "SRC-EA-067-URL-E8CFEB3458D9",
    "SRC-EA-067-URL-D320C481D01C",
    "SRC-EA-067-URL-7C7C7DBBC222",
    "SRC-EA-067-URL-883737C6EFB0",
    "SRC-EA-067-URL-7BBD2F9AC09C",
    "SRC-EA-067-URL-5C9DF414C214",
    "WEBBODY-EA0067-R8-04C80570"
  ],
  "article_sources": [
    {
      "id": "EVD-EA-067-URL-E0E9622033AF",
      "evidence_class": "RESEARCH",
      "title": "1) official product docs | Azure AI Search — Indexers: change and deletion detection | Microsoft |",
      "href": "https://learn.microsoft.com/en-us/azure/search/search-how-to-index-changed-deleted-blobs",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "SOURCE_PACKET_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-EA-067-URL-E0E9622033AF"
      ],
      "supporting_locators": [
        "EVIDENCE TABLE",
        "1) official product docs | Azure AI Search — Indexers: change and deletion detection | Microsoft | https://learn.microsoft.com/en-us/azure/search/search-how-to-index-changed-deleted-blobs"
      ],
      "scope": "EVIDENCE TABLE",
      "provenance_status": "SOURCE_PACKET_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-EA-067-URL-E8CFEB3458D9",
      "evidence_class": "RESEARCH",
      "title": "2) official product docs | Azure AI Search — Indexer overview / scheduling | Microsoft |",
      "href": "https://learn.microsoft.com/en-us/azure/search/search-indexer-overview",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "SOURCE_PACKET_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-EA-067-URL-E8CFEB3458D9"
      ],
      "supporting_locators": [
        "EVIDENCE TABLE",
        "2) official product docs | Azure AI Search — Indexer overview / scheduling | Microsoft | https://learn.microsoft.com/en-us/azure/search/search-indexer-overview"
      ],
      "scope": "EVIDENCE TABLE",
      "provenance_status": "SOURCE_PACKET_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-EA-067-URL-D320C481D01C",
      "evidence_class": "RESEARCH",
      "title": "3) official product docs | Sync your data with your Amazon Bedrock knowledge base | Amazon Web Services |",
      "href": "https://docs.aws.amazon.com/bedrock/latest/userguide/knowledge-base-ingest.html",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "SOURCE_PACKET_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-EA-067-URL-D320C481D01C"
      ],
      "supporting_locators": [
        "EVIDENCE TABLE",
        "3) official product docs | Sync your data with your Amazon Bedrock knowledge base | Amazon Web Services | https://docs.aws.amazon.com/bedrock/latest/userguide/knowledge-base-ingest.html"
      ],
      "scope": "EVIDENCE TABLE",
      "provenance_status": "SOURCE_PACKET_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-EA-067-URL-7C7C7DBBC222",
      "evidence_class": "RESEARCH",
      "title": "4) official product docs | Delete documents from a knowledge base | Amazon Web Services |",
      "href": "https://docs.aws.amazon.com/bedrock/latest/userguide/knowledge-base-delete.html",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "SOURCE_PACKET_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-EA-067-URL-7C7C7DBBC222"
      ],
      "supporting_locators": [
        "EVIDENCE TABLE",
        "4) official product docs | Delete documents from a knowledge base | Amazon Web Services | https://docs.aws.amazon.com/bedrock/latest/userguide/knowledge-base-delete.html"
      ],
      "scope": "EVIDENCE TABLE",
      "provenance_status": "SOURCE_PACKET_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-EA-067-URL-883737C6EFB0",
      "evidence_class": "RESEARCH",
      "title": "5) reproducible benchmark/preprint artifact | StaleBench: A Benchmark for Answer Freshness in Retrieval-Augmented Generation | Karan Singh |",
      "href": "https://zenodo.org/records/20710012",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "SOURCE_PACKET_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-EA-067-URL-883737C6EFB0"
      ],
      "supporting_locators": [
        "EVIDENCE TABLE",
        "5) reproducible benchmark/preprint artifact | StaleBench: A Benchmark for Answer Freshness in Retrieval-Augmented Generation | Karan Singh | https://zenodo.org/records/20710012"
      ],
      "scope": "EVIDENCE TABLE",
      "provenance_status": "SOURCE_PACKET_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-EA-067-URL-7BBD2F9AC09C",
      "evidence_class": "RESEARCH",
      "title": "6) primary preprint | LayerRAG-Bench: A Cross-Layer Reliability Benchmark for Agentic Retrieval-Augmented Generation | Musa Shams |",
      "href": "https://arxiv.org/abs/2607.27353",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "SOURCE_PACKET_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-EA-067-URL-7BBD2F9AC09C"
      ],
      "supporting_locators": [
        "EVIDENCE TABLE",
        "6) primary preprint | LayerRAG-Bench: A Cross-Layer Reliability Benchmark for Agentic Retrieval-Augmented Generation | Musa Shams | https://arxiv.org/abs/2607.27353"
      ],
      "scope": "EVIDENCE TABLE",
      "provenance_status": "SOURCE_PACKET_REFERENCE_PROJECTED"
    },
    {
      "id": "EVD-EA-067-URL-5C9DF414C214",
      "evidence_class": "RESEARCH",
      "title": "7) primary preprint | Risk-Constrained Freshness-Aware Semantic Caching for Open-Web Retrieval-Augmented LLMs | Muhammad Mansoor, Tahir Ahmad, Yeo-Chan Yoon |",
      "href": "https://arxiv.org/abs/2607.04281",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "SOURCE_PACKET_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-EA-067-URL-5C9DF414C214"
      ],
      "supporting_locators": [
        "EVIDENCE TABLE",
        "7) primary preprint | Risk-Constrained Freshness-Aware Semantic Caching for Open-Web Retrieval-Augmented LLMs | Muhammad Mansoor, Tahir Ahmad, Yeo-Chan Yoon | https://arxiv.org/abs/2607.04281"
      ],
      "scope": "EVIDENCE TABLE",
      "provenance_status": "SOURCE_PACKET_REFERENCE_PROJECTED"
    },
    {
      "id": "EA-067-R11-RESEARCH",
      "evidence_class": "RESEARCH",
      "title": "EA-067 監査済みresearch / canonical revision binding",
      "supports": [
        "- VERIFIED_SPEC/INFERENCE: freshnessは少なくとも source mutation time、change-detection time、ingestion start/end、index-visible time、retrieval-visible time、answer-correct time に分解する。単一last_sync timestampでは不足する。\n- VERIFIED_SPEC: incremental indexing/syncは主要providerで実装されるが、方式・対象source・delete semantics・scheduleはprovider/data sourceごとに異なる。存在claimと利用可能条件を分離する。\n- VERIFIED_SPEC/INFERENCE: deletion freshnessはupdate freshnessとは別SLOにする。soft-delete marker、hard delete、tombstone、connector deletion detection、vector-store direct deletionのどれを採るかで残存windowが変わる。\n- RESEARCH: indexが最新でもgenerator/context ordering/cache等によりstale answerが残る可能性がある。よってend-to-end answer freshnessを別測定する。\n- PROPOSAL: `KNOWLEDGE FRESHNESS LEDGER`をCore候補とする。source_object_id, source_version/etag, source_modified_at, mutation_type(add/update/delete), detected_at, ingest_job_id/start/end, indexed_at, retrievable_at, cache_invalidated_at, first_correct_answer_at, stale_window, delete_window, evidence_locatorを保持する。"
      ],
      "limit": "canonical監査・revision bindingの存在を示す。外部ソースcurrentnessが本文またはcontrol-planeで確定していない場合は、currentnessをUNKNOWNのまま保持する。",
      "origin_class": "AUDITED_CANONICAL_RESEARCH_PACKET",
      "source_or_experiment_ids": [
        "COMM-063-EA0067"
      ],
      "supporting_locators": [
        "COMM-063-EA0067",
        "EA-067-R5-LEGACY-MIGRATION-20261008-R1",
        "content/staging/EA-067.md"
      ],
      "scope": "監査済みEA本文・canonical revision・既存Source Graph relationの範囲。欠損currentnessは推測で閉じない。",
      "provenance_status": "AUDIT_PASS_CANONICAL_BINDING"
    },
    {
      "id": "EA-067-R11-WEB-READBACK",
      "evidence_class": "IMPL",
      "title": "EA-067 Web body identity / canonical readback",
      "supports": [
        "canonical staging bodyのSHA-256一致とR9 canonical ledger readbackをProduction projectionの入力として検証する。"
      ],
      "limit": "body identity/readbackは本文claim自体の真偽や外部ソースの最新性を追加保証しない。",
      "checked_at": null,
      "origin_class": "IMPLEMENTATION_READBACK",
      "source_or_experiment_ids": [
        "WEBBODY-EA0067-R8-04C80570"
      ],
      "supporting_locators": [
        "WEBBODY-EA0067-R8-04C80570",
        "content/staging/EA-067.md",
        "EA-067-R5-LEGACY-MIGRATION-20261008-R1"
      ],
      "scope": "R9/R10 canonical readback後のProduction生成時body identity。",
      "provenance_status": "CANONICAL_READBACK_BODY_HASH_MATCH"
    }
  ],
  "research_sections": [
    {
      "heading": "WHY_THIS_MATTERS_TO_UPDATE",
      "text": "RAG/enterprise searchではsourceが更新されても、connector検出、indexer、embedding、vector/keyword index、cache、retrieval、context assemblyの各層が別々の時刻で更新される。したがって「同期成功」「index更新済み」だけでは利用者が最新回答を得ることを保証できない。EA-066のauthorization freshnessとは分離しつつ、同じfreshness SLO体系で扱う必要がある。"
    },
    {
      "heading": "SCOPE_ROUTING",
      "text": "Enterprise AI Core / implementation / provider mapping / benchmark / operability / governance。EA-066 ACL-aware RAG、EA-064 Memory Lifecycleと隣接するが、本Packetはcontent/knowledgeのtemporal freshnessを対象とする。"
    },
    {
      "heading": "CLAIM_OR_RESEARCH_QUESTION",
      "text": "1. Knowledge freshnessはsource→detect→ingest→index→retrieve→answerのend-to-end propertyとして測定すべきか。\n2. update/additionとdeletionは同一同期機構で保証されるか。\n3. incremental sync、full sync、TTL/cache invalidation、CDCの各方式はどこまでfreshnessを保証するか。\n4. indexが最新でもanswerが古いfailureをどう検出するか。"
    },
    {
      "heading": "EVIDENCE TABLE",
      "text": "1) official product docs | Azure AI Search — Indexers: change and deletion detection | Microsoft | https://learn.microsoft.com/en-us/azure/search/search-how-to-index-changed-deleted-blobs | accessed 2026-09-15 | current web docs | locator: change detection / soft delete policy / deletion detection limitations | VERIFIED_SPEC + PRODUCT_VOLATILE | support/limit. Indexerはchanged/new contentを追跡できるが、削除検出にはdata-source-specific policy/configurationが必要で、削除反映を自動一般化できない。\n2) official product docs | Azure AI Search — Indexer overview / scheduling | Microsoft | https://learn.microsoft.com/en-us/azure/search/search-indexer-overview | accessed 2026-09-15 | current web docs | locator: scheduled indexer execution; incremental indexing | VERIFIED_SPEC + PRODUCT_VOLATILE | support/context. scheduled/incremental indexingが存在する一方、freshnessはschedule、change detection、processing timeに依存する。\n3) official product docs | Sync your data with your Amazon Bedrock knowledge base | Amazon Web Services | https://docs.aws.amazon.com/bedrock/latest/userguide/knowledge-base-ingest.html | accessed 2026-09-15 | current Bedrock docs | locator: StartIngestionJob; sync adds/modifies/deletes; incremental sync for supported sources | VERIFIED_SPEC + PRODUCT_VOLATILE | support/limit. Knowledge Baseのsyncは追加・変更・削除を取り込むが、同期実行と完了を経るためsource mutation即時反映とは同義でない。\n4) official product docs | Delete documents from a knowledge base | Amazon Web Services | https://docs.aws.amazon.com/bedrock/latest/userguide/knowledge-base-delete.html | accessed 2026-09-15 | current Bedrock docs | locator: direct ingestion DeleteKnowledgeBaseDocuments; sync caveats | VERIFIED_SPEC + PRODUCT_VOLATILE | support/context. deletionは独立operation/ingestion経路としても扱われ、更新と削除を同一claimに潰せない。\n5) reproducible benchmark/preprint artifact | StaleBench: A Benchmark for Answer Freshness in Retrieval-Augmented Generation | Karan Singh | https://zenodo.org/records/20710012 | 2026-06; accessed 2026-09-15 | current record | locator: catch-up latency/recovery rate; immediate re-indexing experiment across 10 open models | RESEARCH | counter/limit. 即時re-indexでも古い回答が残る例を報告し、index freshness≠answer freshnessを示す。ただしopen-model controlled benchmarkでありenterprise provider保証へ外挿しない。\n6) primary preprint | LayerRAG-Bench: A Cross-Layer Reliability Benchmark for Agentic Retrieval-Augmented Generation | Musa Shams | https://arxiv.org/abs/2607.27353 | 2026-07-29; accessed 2026-09-15 | arXiv 2607.27353 | locator: stale evidence scenario; groundedness false positives | RESEARCH | counter/context. stale evidenceはgroundedness-only評価で見逃され得る。特定providerのfailure rateとしては扱わない。\n7) primary preprint | Risk-Constrained Freshness-Aware Semantic Caching for Open-Web Retrieval-Augmented LLMs | Muhammad Mansoor, Tahir Ahmad, Yeo-Chan Yoon | https://arxiv.org/abs/2607.04281 | 2026-07-05; accessed 2026-09-15 | arXiv 2607.04281 | locator: FreshCache-Bench; time-varying staleness; cache error budgets | RESEARCH | support/context. semantic cacheはfreshnessを別リスクとして扱う必要があり、固定TTLだけではanswer-affecting changeを直接表現しない。"
    },
    {
      "heading": "FINDINGS",
      "text": "- VERIFIED_SPEC/INFERENCE: freshnessは少なくとも source mutation time、change-detection time、ingestion start/end、index-visible time、retrieval-visible time、answer-correct time に分解する。単一last_sync timestampでは不足する。\n- VERIFIED_SPEC: incremental indexing/syncは主要providerで実装されるが、方式・対象source・delete semantics・scheduleはprovider/data sourceごとに異なる。存在claimと利用可能条件を分離する。\n- VERIFIED_SPEC/INFERENCE: deletion freshnessはupdate freshnessとは別SLOにする。soft-delete marker、hard delete、tombstone、connector deletion detection、vector-store direct deletionのどれを採るかで残存windowが変わる。\n- RESEARCH: indexが最新でもgenerator/context ordering/cache等によりstale answerが残る可能性がある。よってend-to-end answer freshnessを別測定する。\n- PROPOSAL: `KNOWLEDGE FRESHNESS LEDGER`をCore候補とする。source_object_id, source_version/etag, source_modified_at, mutation_type(add/update/delete), detected_at, ingest_job_id/start/end, indexed_at, retrievable_at, cache_invalidated_at, first_correct_answer_at, stale_window, delete_window, evidence_locatorを保持する。"
    },
    {
      "heading": "LIMITS / COUNTEREVIDENCE / FAILURE ENVELOPE",
      "text": "- connectorが「incremental」を称しても全source type・全metadata・ACL・delete eventの完全CDCを意味しない。\n- full re-indexはfreshnessを改善し得るが、処理中のmixed-version window、費用、latency、availabilityを生む。\n- TTLは時間経過による再取得契機であり、source changeを即時検出するCDCではない。\n- cache invalidation成功はindex更新を証明せず、index更新成功はanswer freshnessを証明しない。\n- StaleBench/LayerRAG-Benchはfailure classの存在証拠であり、特定商用tenantのfailure probabilityではない。"
    },
    {
      "heading": "AVAILABILITY_GATES",
      "text": "provider / connector type / source API capability / change token・timestamp・soft-delete fieldの有無 / scheduler minimum interval / direct-ingestion API / vector store / region / plan / permissions。PRODUCT_VOLATILEとしてfuture update時に再確認する。"
    },
    {
      "heading": "IMPLEMENTATION_STATE",
      "text": "Azure AI Search: scheduled indexers、change detection、data-source-specific deletion detectionを実装。AWS Bedrock Knowledge Bases: ingestion/syncおよびdirect document deletion経路を実装。provider横断で同一freshness contractが存在することはUNKNOWN。"
    },
    {
      "heading": "BENCHMARK_OR_EVAL",
      "text": "TEST_PROFILE（未実施）: synthetic corpusにK=v1を投入しbaseline確認→sourceをv2へ更新→t0からsource/index/retrieval/answerをpoll→同documentをdelete→同様に残存をpoll。`T_detect`, `T_ingest`, `T_index_visible`, `T_retrieval_fresh`, `T_answer_fresh`, `T_delete_retrieval`, `T_delete_answer`を個別記録。cache ON/OFF、incremental/full、conflicting v1/v2 coexistenceを分ける。正解判定は可能な限りexact controlled factを使いLLM judge依存を避ける。"
    },
    {
      "heading": "SECURITY / PRIVACY / AUDIT / COST / OPERABILITY",
      "text": "AUDIT: mutation→sync job→index version→query evidenceを相関可能にする。OPERABILITY: freshness SLO違反をjob successとは別alertにする。COST: refresh頻度を上げるほどconnector/API/embedding/indexing/cache miss costが増えるため、data class別SLOが必要。SECURITY: EA-066のACL revocation freshnessとcontent freshnessを別metricで測る。削除済み機密情報のstale retrievalはsecurity/privacy incidentになり得る。"
    },
    {
      "heading": "SAFE_WORDING_FOR_FUTURE_BOOK",
      "text": "「RAGの鮮度は、データソースの更新時刻やインデックス同期成功だけでは判定できない。更新・削除が検出され、検索可能な表現へ反映され、キャッシュや検索結果を経て、最終回答が新しい状態へ追随するまでの各段階を測定する。incremental sync、TTL、CDC、再indexはそのための手段であり、それぞれの製品・connectorが同じ反映保証を持つとは限らない。」"
    },
    {
      "heading": "UPDATE_WORK_HANDOFF",
      "text": "1. future update時にAzure/AWS/Google等のcurrent connector別sync/deletion semanticsとminimum intervalを再確認。\n2. EA-066の`revocation-to-enforcement latency`と本EAの`mutation-to-answer latency`を同一Freshness SLO章で並置するが混同しない。\n3. KNOWLEDGE FRESHNESS LEDGER採用可否をCoreで判断。\n4. high-risk corpusではupdate/delete canaryを定期実行し、job successではなくretrieval/answer layerまで検証する設計を候補化。"
    },
    {
      "heading": "UNKNOWN",
      "text": "provider横断で標準化されたend-to-end freshness metadata/schema、各managed connectorの実tenantでのp95/p99 propagation latency、hard-delete後の全cache/derived artifact残存window。STOP_CONDITION: update/add/delete各mutationについてsource→answerまでのdocumented semanticsとsafe synthetic measurementが揃い、provider/connector別availability gateが更新時点で再確認できた時点。\n\n---"
    }
  ],
  "human_url": "/tools/ea-067-knowledge-freshness-engineering",
  "markdown_url": "/agent/article.md?path=%2Ftools%2Fea-067-knowledge-freshness-engineering",
  "json_url": "/agent/article.json?path=%2Ftools%2Fea-067-knowledge-freshness-engineering"
}