Save To Graph

使用方法

# 標準実行（全未処理 JSON を投入）
/save-to-graph

# 特定コマンドソースのみ
/save-to-graph --source dr-stock

# 特定ファイルのみ
/save-to-graph --file .tmp/graph-queue/ca-eval/gq-20260317120000-a1b2.json

# ドライラン
/save-to-graph --dry-run

# 処理済みファイルを保持
/save-to-graph --keep

パラメータ

前提条件

1. Neo4j が起動していること（quants-neo4j: bolt://localhost:7687）
2. 初回セットアップが完了していること（制約・インデックス作成）
3. graph-queue JSON が存在すること（scripts/emit_graph_queue.py で生成）

Phase 1: キュー検出・検証

ステップ 1.1: Neo4j 接続確認

neo4j-cypher MCP の read_query で接続テスト:

RETURN 'connection_ok' AS status

ステップ 1.2: 未処理ファイル検出

.tmp/graph-queue/ 配下の gq-*.json を検出。

ステップ 1.3: JSON スキーマ検証

必須キー:

required_keys = {
    "schema_version", "queue_id", "created_at", "command_source",
    "sources", "entities", "claims", "facts", "topics",
    "authors", "financial_datapoints", "fiscal_periods", "insights",
    "relations",
}

# v2.0 追加キー（schema_version "1.0" では省略可、"2.0" では必須）
v2_keys = {
    "anomalies", "performance_evidences", "market_regimes", "data_requirements",
}

Phase 2: ノード投入（MERGE）

投入順序は依存関係に基づく: Topic → Entity → FiscalPeriod → Source → Author → Fact → Claim → FinancialDataPoint → Insight → Method → Anomaly → PerformanceEvidence → MarketRegime → DataRequirement

Topic ノード MERGE

MERGE (t:Topic {topic_id: $topic_id})
SET t.name = $name, t.category = $category, t.is_meta = COALESCE($is_meta, false)

Entity ノード MERGE

MERGE (e:Entity {entity_id: $entity_id})
SET e.name = $name, e.entity_type = $entity_type, e.ticker = $ticker

FiscalPeriod ノード MERGE

MERGE (fp:FiscalPeriod {period_id: $period_id})
SET fp.period_type = $period_type, fp.period_label = $period_label

Source ノード MERGE

MERGE (s:Source {source_id: $source_id})
SET s.url = $url, s.title = $title, s.source_type = $source_type,
    s.command_source = $command_source,
    s.published_at = COALESCE($published_at, $fetched_at, s.published_at),
    s.fetched_at = COALESCE($fetched_at, s.fetched_at)

Author ノード MERGE

MERGE (a:Author {author_id: $author_id})
SET a.name = $name, a.author_type = $author_type

Fact ノード MERGE

MERGE (f:Fact {fact_id: $fact_id})
SET f.content = $content, f.fact_type = $fact_type,
    f.created_at = datetime($created_at)

Claim ノード MERGE

MERGE (c:Claim {claim_id: $claim_id})
SET c.content = $content, c.claim_type = $claim_type,
    c.sentiment = $sentiment,
    c.confidence = COALESCE($confidence, 'medium'),
    c.created_at = datetime($created_at)

FinancialDataPoint ノード MERGE

MERGE (dp:FinancialDataPoint {datapoint_id: $datapoint_id})
SET dp.metric_name = $metric_name, dp.value = $value,
    dp.unit = $unit, dp.is_estimate = $is_estimate

Insight ノード MERGE

MERGE (i:Insight {insight_id: $insight_id})
SET i.content = $content, i.insight_type = $insight_type,
    i.generated_at = datetime($generated_at)

Method ノード MERGE

MERGE (m:Method {method_id: $method_id})
SET m.name = $name, m.method_type = $method_type, m.description = $description

Anomaly ノード MERGE

MERGE (a:Anomaly {anomaly_id: $anomaly_id})
SET a.name = $name, a.anomaly_type = $anomaly_type,
    a.persistence = $persistence, a.created_at = datetime()

PerformanceEvidence ノード MERGE

MERGE (pe:PerformanceEvidence {evidence_id: $evidence_id})
SET pe.metric_name = $metric_name, pe.value = $value,
    pe.unit = $unit, pe.benchmark = $benchmark, pe.market = $market,
    pe.is_oos = $is_oos, pe.created_at = datetime()

MarketRegime ノード MERGE

MERGE (mr:MarketRegime {regime_id: $regime_id})
SET mr.name = $name, mr.regime_type = $regime_type,
    mr.description = $description, mr.created_at = datetime()

DataRequirement ノード MERGE

MERGE (dr:DataRequirement {data_req_id: $data_req_id})
SET dr.name = $name, dr.data_type = $data_type,
    dr.frequency = $frequency, dr.availability = $availability,
    dr.created_at = datetime()

Phase 3a: ファイル内リレーション投入

graph-queue JSON の relations オブジェクトから各リレーションを MERGE:

• tagged[] → TAGGED (Source → Topic)
• makes_claim[] → MAKES_CLAIM (Source → Claim)
• about[] → ABOUT (Claim → Entity)
• states_fact[] → STATES_FACT (Source → Fact)
• has_datapoint[] → HAS_DATAPOINT (Entity → FinancialDataPoint)
• for_period[] → FOR_PERIOD (FinancialDataPoint → FiscalPeriod)
• relates_to[] → RELATES_TO (Fact → Entity)
• supported_by[] → SUPPORTED_BY (Claim → Fact)
• authored_by[] → AUTHORED_BY (Source → Author)
• uses_method[] → USES_METHOD (Source → Method)
• exploits[] → EXPLOITS (Method → Anomaly)
• evaluates[] → EVALUATES (PerformanceEvidence → Method)
• quantified_by[] → QUANTIFIED_BY (Claim → PerformanceEvidence)
• effective_in[] → EFFECTIVE_IN (Method → MarketRegime)
• requires_data[] → REQUIRES_DATA (Method → DataRequirement)
• explained_by[] → EXPLAINED_BY (Anomaly → Claim)
• measured_in[] → MEASURED_IN (PerformanceEvidence → MarketRegime)
• extends_method[] → EXTENDS_METHOD (Method → Method)
• combined_with[] → COMBINED_WITH (Method → Method)
• competes_with[] → COMPETES_WITH (Method → Method)
• contradicts[] → CONTRADICTS (Claim → Claim)
• cites[] → CITES (Source → Source)
• coauthored_with[] → COAUTHORED_WITH (Author → Author)
• subtopic_of[] → SUBTOPIC_OF (Topic → Topic)
• affiliated_with[] → AFFILIATED_WITH (Author → Entity)

Phase 3b: クロスファイルリレーション

TAGGED カテゴリマッチング

UNWIND $source_ids AS sid
MATCH (s:Source {source_id: sid})
MATCH (t:Topic {category: s.category})
MERGE (s)-[:TAGGED]->(t)

ABOUT コンテンツマッチング

UNWIND $claim_ids AS cid
MATCH (c:Claim {claim_id: cid})
MATCH (e:Entity)
WHERE size(e.name) >= 2 AND c.content CONTAINS e.name
MERGE (c)-[:ABOUT]->(e)

Phase 4: 完了処理

冪等性の保証

1. ノード投入: MERGE はノード存在時に更新、不在時に作成
2. リレーション投入: MERGE はリレーション存在時に何もしない
3. ID の決定論性: 全 ID は入力データから決定論的に生成（UUID5 / SHA-256）

エラーハンドリング

環境変数

対応コマンドソース

関連リソース