Copy execution transparency (Phase 3)

Per-copy execution facts — latency realized slippage fill vs failure — captured ทุก copy attempt surfaced เป็น per-profile transparency report Off by default; enable ด้วย App:Copy:TransparencyEnabled=true เมื่อ off copy engine byte-for-byte ไม่เปลี่ยนแปลง: host emits ไป no-op sink nothing written

How มันworks

CopyEngineHost ──Record(fact)──▶ ICopyEventSink
                                   │
             (transparency off) NullCopyEventSink   → discards (default; zero hot-path cost)
             (transparency on)  ChannelCopyEventSink → bounded in-memory channel (DropOldest)
                                   │
                                   ▼
                          CopyExecutionDrainer (BackgroundService)
                                   │  batches ทุก App drain interval
                                   ▼
                          CopyExecution append-only table  ◀── GET /api/copy/profiles/{id}/transparency

• Hot path stays free ของ I/O Host calls ICopyEventSink.Record(...) — non-blocking never-throwing enqueue ไม่เคย awaits ไม่เคยtouches DB ไม่เคยblocks order execution
• Loss preferred over back-pressure Channel bounded (CopyExecutionChannelCapacity) ด้วย DropOldest: if DB drainer stalls oldest transparency rows dropped แทน delay copy transparency = best-effort telemetry ไม่ trading dependency
• Out-of-band persistence CopyExecutionDrainer drains channel ใน batches (CopyExecutionDrainBatchSize) on CopyExecutionDrainInterval writes CopyExecution rows ผ่าน scoped DataContext final flush on shutdown
• Facts ไม่ใช่ commands CopyExecution = append-only log (like InstanceLog/AuditLog) ไม่ aggregate read model queries มัน directly (CQRS-lite) aggregates ใน memory

What recorded

One CopyExecutionRecord per copy attempt on one destination:

Kind	When	Carries
Opened	copy order placed	symbol side wire volume master price realized slippage (points) latency (ms)
Failed	copy open threw/rejected	symbol side master volume/price latency failure reason (exception type)

(Closed/Skipped/Reconciled exist ใน enum สำหรับ future expansion)

The report

GET /api/copy/profiles/{id}/transparency (owner-scoped) returns over most recent 500 facts:

• Summary — total opened failed fill rate average latency (ms) average slippage (points)
• Recent — raw recent facts (destination source position symbol side volume master price slippage latency reason timestamp)

Configuration (App:Copy)

Setting	Default	Effect
TransparencyEnabled	false	Turn per-copy fact capture + drainer on สำหรับ node

Channel capacity drain batch size drain interval = CopyDefaults constants (CopyExecutionChannelCapacity / CopyExecutionDrainBatchSize / CopyExecutionDrainInterval)

Tests

• Unit (CopyTransparencyTests) — successful open emits Opened fact ด้วย right symbol/side/volume/latency; rejected open emits Failed fact ด้วย reason driven ผ่าน capturing sink
• Integration (CopyExecutionDrainerTests real Postgres) — drainer persists buffered facts ไป CopyExecution log; empty sink writes nothing
• DST — host change fire-and-forget ด้วย no-op default sink ดังนั้น deterministic copy stress suite stays green (23/23)