docs: update ClickHouse schema (http_logs_raw + http_logs)

- README.md: documenter les deux tables (raw + enrichie)
- architecture.yml: décrire le schema complet avec colonnes matérialisées
- Table http_logs_raw: ingestion JSON brut (colonne raw_json unique)
- Table http_logs: extraction des champs via DEFAULT JSONExtract*

Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>

README.md

@@ -194,18 +194,81 @@ Tous les champs des sources A et B sont fusionnés au même niveau. Les champs d
 
 ## Schema ClickHouse
 
+Le service utilise deux tables ClickHouse :
+
+### Table brute (`http_logs_raw`)
+
+Table d'ingestion qui stocke le log corrélé brut au format JSON :
+
 ```sql
-CREATE TABLE correlated_logs_http_network (
-    timestamp DateTime64(9),
-    src_ip String,
-    src_port UInt32,
-    dst_ip String,
-    dst_port UInt32,
-    correlated UInt8,
-    orphan_side String,
-    fields JSON
-) ENGINE = MergeTree()
-ORDER BY (timestamp, src_ip, src_port);
+CREATE TABLE http_logs_raw
+(
+    raw_json String
+)
+ENGINE = MergeTree
+ORDER BY tuple();
+```
+
+**Format d'insertion :** Le service envoie chaque log corrélé sérialisé en JSON dans la colonne `raw_json` :
+
+```sql
+INSERT INTO http_logs_raw (raw_json) FORMAT JSONEachRow
+{"raw_json":"{\"timestamp\":\"2024-01-01T12:00:00Z\",\"src_ip\":\"192.168.1.1\",\"correlated\":true,...}"}
+```
+
+### Table enrichie (`http_logs`)
+
+Vue matérialisée qui extrait les champs du JSON pour l'analyse :
+
+```sql
+CREATE TABLE http_logs
+(
+    raw_json String,
+
+    -- champs de base
+    time_str String DEFAULT JSONExtractString(raw_json, 'time'),
+    timestamp_str String DEFAULT JSONExtractString(raw_json, 'timestamp'),
+    time DateTime DEFAULT parseDateTimeBestEffort(time_str),
+    log_date Date DEFAULT toDate(time),
+
+    src_ip IPv4 DEFAULT toIPv4(JSONExtractString(raw_json, 'src_ip')),
+    src_port UInt16 DEFAULT toUInt16(JSONExtractUInt(raw_json, 'src_port')),
+    dst_ip IPv4 DEFAULT toIPv4(JSONExtractString(raw_json, 'dst_ip')),
+    dst_port UInt16 DEFAULT toUInt16(JSONExtractUInt(raw_json, 'dst_port')),
+
+    correlated UInt8 DEFAULT JSONExtractBool(raw_json, 'correlated'),
+    keepalives UInt16 DEFAULT toUInt16(JSONExtractUInt(raw_json, 'keepalives')),
+    method LowCardinality(String) DEFAULT JSONExtractString(raw_json, 'method'),
+    scheme LowCardinality(String) DEFAULT JSONExtractString(raw_json, 'scheme'),
+    host LowCardinality(String) DEFAULT JSONExtractString(raw_json, 'host'),
+    path String DEFAULT JSONExtractString(raw_json, 'path'),
+    query String DEFAULT JSONExtractString(raw_json, 'query'),
+    http_version LowCardinality(String) DEFAULT JSONExtractString(raw_json, 'http_version'),
+    orphan_side LowCardinality(String) DEFAULT JSONExtractString(raw_json, 'orphan_side'),
+
+    -- champs « presque toujours là »
+    a_timestamp UInt64 DEFAULT JSONExtractUInt(raw_json, 'a_timestamp'),
+    b_timestamp UInt64 DEFAULT JSONExtractUInt(raw_json, 'b_timestamp'),
+    conn_id String DEFAULT JSONExtractString(raw_json, 'conn_id'),
+    ip_meta_df UInt8 DEFAULT JSONExtractBool(raw_json, 'ip_meta_df'),
+    ip_meta_id UInt32 DEFAULT JSONExtractUInt(raw_json, 'ip_meta_id'),
+    ip_meta_total_length UInt32 DEFAULT JSONExtractUInt(raw_json, 'ip_meta_total_length'),
+    ip_meta_ttl UInt8 DEFAULT JSONExtractUInt(raw_json, 'ip_meta_ttl'),
+    tcp_meta_options LowCardinality(String) DEFAULT JSONExtractString(raw_json, 'tcp_meta_options'),
+    tcp_meta_window_size UInt32 DEFAULT JSONExtractUInt(raw_json, 'tcp_meta_window_size'),
+    syn_to_clienthello_ms Int32 DEFAULT toInt32(JSONExtractInt(raw_json, 'syn_to_clienthello_ms')),
+    tls_version LowCardinality(String) DEFAULT JSONExtractString(raw_json, 'tls_version'),
+    tls_sni LowCardinality(String) DEFAULT JSONExtractString(raw_json, 'tls_sni'),
+    ja3 String DEFAULT JSONExtractString(raw_json, 'ja3'),
+    ja3_hash String DEFAULT JSONExtractString(raw_json, 'ja3_hash'),
+    ja4 String DEFAULT JSONExtractString(raw_json, 'ja4'),
+
+    -- tous les autres champs JSON (headers dynamiques etc.)
+    extra JSON DEFAULT raw_json
+)
+ENGINE = MergeTree
+PARTITION BY toYYYYMM(log_date)
+ORDER BY (log_date, dst_ip, src_ip, time);
 ```
 
 ## Tests