agentworkshop 0.5.9 → 0.6.0

package/server/services/workshop/daq/png-enc.ts

@@ -0,0 +1,71 @@
+/**
+ * 最小 PNG 编码器(zlib 内建,零外部依赖)—— mock CCD 驱动与 thumbnail 处理器共用。
+ *
+ * 支持 8-bit 灰度(color type 0)与 8-bit RGB(color type 2);scanline filter 0
+ * (None)—— 压缩率略逊于 Paeth,但编码零卷积开销,采集管线内可忽略。
+ */
+import { deflateSync } from 'node:zlib'
+
+const CRC_TABLE = (() => {
+  const t = new Uint32Array(256)
+  for (let n = 0; n < 256; n++) {
+    let c = n
+    for (let k = 0; k < 8; k++) c = c & 1 ? 0xedb88320 ^ (c >>> 1) : c >>> 1
+    t[n] = c >>> 0
+  }
+  return t
+})()
+
+function crc32(buf: Buffer): number {
+  let c = 0xffffffff
+  for (let i = 0; i < buf.length; i++) c = CRC_TABLE[(c ^ buf[i]!) & 0xff]! ^ (c >>> 8)
+  return (c ^ 0xffffffff) >>> 0
+}
+
+function chunk(type: string, data: Buffer): Buffer {
+  const len = Buffer.alloc(4)
+  len.writeUInt32BE(data.length)
+  const body = Buffer.concat([Buffer.from(type, 'ascii'), data])
+  const crc = Buffer.alloc(4)
+  crc.writeUInt32BE(crc32(body))
+  return Buffer.concat([len, body, crc])
+}
+
+/** 灰度/RGB 原始像素 → PNG Buffer。gray 模式 pixels.length === w*h;rgb === w*h*3(RGB 交错)。 */
+export function encodePng(width: number, height: number, pixels: Uint8Array, rgb = false): Buffer {
+  const colorType = rgb ? 2 : 0
+  const bpp = rgb ? 3 : 1
+  const stride = width * bpp
+  // 每行前置 filter byte 0(None)
+  const raw = Buffer.alloc((stride + 1) * height)
+  for (let y = 0; y < height; y++) {
+    raw[y * (stride + 1)] = 0
+    pixels.subarray(y * stride, (y + 1) * stride).forEach((v, i) => {
+      raw[y * (stride + 1) + 1 + i] = v
+    })
+  }
+  const ihdr = Buffer.alloc(13)
+  ihdr.writeUInt32BE(width, 0)
+  ihdr.writeUInt32BE(height, 4)
+  ihdr[8] = 8 // bit depth
+  ihdr[9] = colorType
+  return Buffer.concat([
+    Buffer.from([0x89, 0x50, 0x4e, 0x47, 0x0d, 0x0a, 0x1a, 0x0a]),
+    chunk('IHDR', ihdr),
+    chunk('IDAT', deflateSync(raw, { level: 6 })),
+    chunk('IEND', Buffer.alloc(0)),
+  ])
+}
+
+/** 最近邻缩放(灰度;thumbnail 处理器用;像素抽点,零插值开销) */
+export function resizeGray(src: Uint8Array, w: number, h: number, tw: number, th: number): Uint8Array {
+  const out = new Uint8Array(tw * th)
+  for (let y = 0; y < th; y++) {
+    const sy = Math.min(h - 1, Math.floor((y * h) / th))
+    for (let x = 0; x < tw; x++) {
+      const sx = Math.min(w - 1, Math.floor((x * w) / tw))
+      out[y * tw + x] = src[sy * w + sx]!
+    }
+  }
+  return out
+}

package/server/services/workshop/daq/storage/sqlite.adapter.ts

@@ -50,10 +50,30 @@ export class SqliteTimeSeriesAdapter implements TsdbPort {
       CREATE INDEX IF NOT EXISTS idx_daq_tag ON daq_samples (product_id, recipe_id, ts_ms DESC);
       CREATE INDEX IF NOT EXISTS idx_daq_run ON daq_samples (run_id, ts_ms DESC);
     `)
+    // 帧表(v2:向量/图像;meta/metrics JSON 序列化 TEXT,与 Timescale 契约对齐)
+    this.db.exec(`
+      CREATE TABLE IF NOT EXISTS daq_frames (
+        node_id TEXT    NOT NULL,
+        ts_ms   INTEGER NOT NULL,
+        kind    TEXT    NOT NULL,
+        template_key TEXT,
+        device_binding_id TEXT,
+        line_id TEXT, product_id TEXT, recipe_id TEXT, run_id TEXT,
+        points  INTEGER NOT NULL DEFAULT 0,
+        meta    TEXT    NOT NULL DEFAULT '{}',
+        metrics TEXT    NOT NULL DEFAULT '{}',
+        PRIMARY KEY (node_id, ts_ms)
+      )
+    `)
+    this.db.exec('CREATE INDEX IF NOT EXISTS idx_daq_frames_node_ts ON daq_frames (node_id, ts_ms DESC)')
     // 保留窗口:启动清一次 + 后台周期任务(30min,防长会话磁盘无限涨)
     this.sweepRetention()
+    this.sweepFrameRetention()
     if (!this.retentionTimer) {
-      this.retentionTimer = setInterval(() => this.sweepRetention(), 30 * 60_000)
+      this.retentionTimer = setInterval(() => {
+        this.sweepRetention()
+        this.sweepFrameRetention()
+      }, 30 * 60_000)
       this.retentionTimer.unref?.()
     }
   }
@@ -190,6 +210,70 @@ export class SqliteTimeSeriesAdapter implements TsdbPort {
     return out
   }
 
+  // ===== 帧(daq_frames;与 Timescale 同契约)=====
+
+  async writeFrames(rows: import('./tsdb-port').DaqFrameRow[]): Promise<void> {
+    if (!this.db || rows.length === 0) return
+    const ins = this.db.prepare(
+      'INSERT OR IGNORE INTO daq_frames (node_id, ts_ms, kind, template_key, device_binding_id, line_id, product_id, recipe_id, run_id, points, meta, metrics) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)',
+    )
+    this.db.exec('BEGIN')
+    try {
+      for (const r of rows) {
+        ins.run(r.nodeId, r.tsMs, r.kind, r.templateKey ?? null, r.deviceBindingId ?? null, r.lineId ?? null, r.productId ?? null, r.recipeId ?? null, r.runId ?? null, r.points, JSON.stringify(r.meta ?? {}), JSON.stringify(r.metrics ?? {}))
+      }
+      this.db.exec('COMMIT')
+    }
+    catch (err) {
+      try {
+        this.db.exec('ROLLBACK')
+      }
+      catch { /* 事务已终结 */ }
+      throw err
+    }
+  }
+
+  async queryFrames(nodeId: string, opts: import('./tsdb-port').DaqFrameQueryOpts): Promise<import('./tsdb-port').DaqFrameRecord[]> {
+    if (!this.db) return []
+    const limit = Math.min(opts.limit ?? 100, 1000)
+    const rows = this.db.prepare(`
+      SELECT ts_ms, kind, meta, metrics, device_binding_id, line_id, product_id, recipe_id, run_id
+      FROM daq_frames
+      WHERE node_id = :id AND (:kind IS NULL OR kind = :kind) AND ts_ms >= :from AND ts_ms <= :to
+      ORDER BY ts_ms DESC LIMIT :lim
+    `).all({
+      ':id': nodeId,
+      ':kind': opts.kind ?? null,
+      ':from': opts.fromMs ?? 0,
+      ':to': opts.toMs ?? Date.now(),
+      ':lim': limit,
+    }) as Array<{ ts_ms: number, kind: string, meta: string, metrics: string, device_binding_id: string | null, line_id: string | null, product_id: string | null, recipe_id: string | null, run_id: string | null }>
+    return rows.map(r => ({
+      at: Number(r.ts_ms),
+      kind: String(r.kind) as 'vector' | 'image',
+      points: r.kind === 'vector' ? pointsFromMeta(asObject(r.meta)) : undefined,
+      metrics: asObject(r.metrics) as Record<string, number>,
+      meta: asObject(r.meta),
+      deviceBindingId: r.device_binding_id,
+      lineId: r.line_id,
+      productId: r.product_id,
+      recipeId: r.recipe_id,
+      runId: r.run_id,
+    }))
+  }
+
+  /** 帧保留期清理(与样本同拍;帧保留期独立可配 DAQ_FRAME_RETENTION_H,默认 720h) */
+  private sweepFrameRetention(): void {
+    if (!this.db) return
+    const cutoff = Date.now() - Number(process.env.DAQ_FRAME_RETENTION_H ?? 720) * 3600_000
+    try {
+      this.db.prepare('DELETE FROM daq_frames WHERE ts_ms < ?').run(cutoff)
+    }
+    catch (err) {
+      log.error('[daq-tsdb] 帧保留期清理失败:', err instanceof Error ? err.message : err)
+    }
+  }
+
   close(): Promise<void> | void {
     this.sweepRetentionTimerStop()
     this.db?.close()
@@ -203,3 +287,22 @@ export class SqliteTimeSeriesAdapter implements TsdbPort {
     }
   }
 }
+
+/** 向量点列从 meta JSON 还原(非数组/越界防御) */
+function pointsFromMeta(meta: Record<string, unknown>): number[] | undefined {
+  const p = meta.points
+  if (!Array.isArray(p) || p.length === 0 || p.length > 4096) return undefined
+  return p.every(x => Number.isFinite(Number(x))) ? p.map(Number) : undefined
+}
+
+function asObject(v: unknown): Record<string, unknown> {
+  if (typeof v === 'string') {
+    try {
+      return JSON.parse(v) as Record<string, unknown>
+    }
+    catch {
+      return {}
+    }
+  }
+  return (v && typeof v === 'object' ? v : {}) as Record<string, unknown>
+}

package/server/services/workshop/daq/storage/timescale.adapter.ts

@@ -19,6 +19,8 @@ export class TimescaleAdapter implements TsdbPort {
   private retentionTimer: NodeJS.Timeout | null = null
   /** 保留期(默认 7 天,DAQ_TS_RETENTION_H 可调) */
   private readonly retentionH = Number(process.env.DAQ_TS_RETENTION_H ?? 168)
+  /** 帧保留期(默认 30 天,DAQ_FRAME_RETENTION_H 可调;图像元数据体量远小于标量) */
+  private readonly frameRetentionH = Number(process.env.DAQ_FRAME_RETENTION_H ?? 720)
 
   constructor(private readonly url: string) {}
 
@@ -49,6 +51,11 @@ export class TimescaleAdapter implements TsdbPort {
       await this.pool.query(ddl).catch(() => {})
     }
     await this.pool.query(`CREATE INDEX IF NOT EXISTS idx_daq_samples_tag ON daq_samples (product_id, recipe_id, ts DESC)`).catch(() => {})
+    // 帧表(v2:向量/图像元数据;像素 blob 在对象存储,JSONB 承载点列与派生指标;
+    // DDL 为静态字面量,零外部输入)
+    await this.pool.query('CREATE TABLE IF NOT EXISTS daq_frames (node_id text NOT NULL, ts timestamptz NOT NULL, kind text NOT NULL, template_key text, device_binding_id text, line_id text, product_id text, recipe_id text, run_id text, points integer NOT NULL DEFAULT 0, meta jsonb NOT NULL DEFAULT \'{}\'::jsonb, metrics jsonb NOT NULL DEFAULT \'{}\'::jsonb, PRIMARY KEY (node_id, ts))').catch(() => {})
+    await this.pool.query(`SELECT create_hypertable('daq_frames', 'ts', if_not_exists => TRUE, migrate_data => TRUE)`).catch(() => {})
+    await this.pool.query(`CREATE INDEX IF NOT EXISTS idx_daq_frames_node_ts ON daq_frames (node_id, ts DESC)`).catch(() => {})
     // 保留期:drop_chunks 周期执行(30min;未装扩展时静默跳过)
     void this.sweepRetention()
     if (!this.retentionTimer) {
@@ -64,6 +71,10 @@ export class TimescaleAdapter implements TsdbPort {
         `SELECT drop_chunks('daq_samples', older_than => now() - ($1 || ' hours')::interval)`,
         [this.retentionH],
       )
+      await this.pool.query(
+        `SELECT drop_chunks('daq_frames', older_than => now() - ($1 || ' hours')::interval)`,
+        [this.frameRetentionH],
+      )
     }
     catch { /* 非 hypertable(扩展缺失)→ 静默 */ }
   }
@@ -110,7 +121,7 @@ export class TimescaleAdapter implements TsdbPort {
        ORDER BY ts DESC LIMIT $4`,
       [nodeId, new Date(opts.fromMs ?? 0).toISOString(), new Date(opts.toMs ?? Date.now()).toISOString(), limit],
     )
-    return rows.map(r => ({ at: Date.parse(r.ts), value: Number(r.value), state: String(r.state) }))
+    return rows.map(r => ({ at: tsToMs(r.ts), value: Number(r.value), state: String(r.state) }))
   }
 
   async queryTagged(q: import('./tsdb-port').TsdbTagQuery): Promise<Map<string, import('./tsdb-port').TsdbPoint[]>> {
@@ -154,7 +165,7 @@ export class TimescaleAdapter implements TsdbPort {
       )
       for (const r of rows as Array<{ node_id: string, b_at: string, avg: string, min: string, max: string, cnt: string }>) {
         const list = out.get(String(r.node_id)) ?? []
-        list.push({ at: Date.parse(r.b_at), avg: Number(r.avg), min: Number(r.min), max: Number(r.max), cnt: Number(r.cnt) })
+        list.push({ at: tsToMs(r.b_at), avg: Number(r.avg), min: Number(r.min), max: Number(r.max), cnt: Number(r.cnt) })
         out.set(String(r.node_id), list)
       }
       return out
@@ -165,7 +176,7 @@ export class TimescaleAdapter implements TsdbPort {
     )
     for (const r of rows as Array<{ node_id: string, ts: string, value: number }>) {
       const list = out.get(String(r.node_id)) ?? []
-      list.push({ at: Date.parse(r.ts), value: Number(r.value) })
+      list.push({ at: tsToMs(r.ts), value: Number(r.value) })
       out.set(String(r.node_id), list)
     }
     return out
@@ -178,8 +189,88 @@ export class TimescaleAdapter implements TsdbPort {
       `SELECT DISTINCT ON (node_id) node_id, ts, value, state FROM daq_samples ORDER BY node_id, ts DESC`,
     )
     for (const r of rows) {
-      out.set(String(r.node_id), { nodeId: String(r.node_id), tsMs: Date.parse(r.ts), value: Number(r.value), state: String(r.state) })
+      out.set(String(r.node_id), { nodeId: String(r.node_id), tsMs: tsToMs(r.ts), value: Number(r.value), state: String(r.state) })
     }
     return out
   }
+
+  // ===== 帧(daq_frames;元数据+JSONB,像素在对象存储)=====
+
+  async writeFrames(rows: import('./tsdb-port').DaqFrameRow[]): Promise<void> {
+    if (!this.pool || rows.length === 0) return
+    // 固定占位符逐行插入 + 客户端事务(帧量级 ≤ 每节点每周期 1 条,无需动态 VALUES)
+    const client = await this.pool.connect()
+    try {
+      await client.query('BEGIN')
+      for (const r of rows) {
+        await client.query(
+          `INSERT INTO daq_frames (node_id, ts, kind, template_key, device_binding_id, line_id, product_id, recipe_id, run_id, points, meta, metrics)
+           VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11::jsonb, $12::jsonb)
+           ON CONFLICT (node_id, ts) DO NOTHING`,
+          [
+            r.nodeId, new Date(r.tsMs).toISOString(), r.kind, r.templateKey ?? null, r.deviceBindingId ?? null,
+            r.lineId ?? null, r.productId ?? null, r.recipeId ?? null, r.runId ?? null,
+            r.points, JSON.stringify(r.meta ?? {}), JSON.stringify(r.metrics ?? {}),
+          ],
+        )
+      }
+      await client.query('COMMIT')
+    }
+    catch (err) {
+      await client.query('ROLLBACK').catch(() => {})
+      throw err
+    }
+    finally {
+      client.release()
+    }
+  }
+
+  async queryFrames(nodeId: string, opts: import('./tsdb-port').DaqFrameQueryOpts): Promise<import('./tsdb-port').DaqFrameRecord[]> {
+    if (!this.pool) return []
+    const limit = Math.min(opts.limit ?? 100, 1000)
+    const { rows } = await this.pool.query(
+      `SELECT ts, kind, points, meta, metrics, device_binding_id, line_id, product_id, recipe_id, run_id
+       FROM daq_frames WHERE node_id = $1 AND ($2::text IS NULL OR kind = $2) AND ts >= $3 AND ts <= $4
+       ORDER BY ts DESC LIMIT $5`,
+      [nodeId, opts.kind ?? null, new Date(opts.fromMs ?? 0).toISOString(), new Date(opts.toMs ?? Date.now()).toISOString(), limit],
+    )
+    return rows.map(r => ({
+      at: tsToMs(r.ts),
+      kind: String(r.kind) as 'vector' | 'image',
+      points: String(r.kind) === 'vector' ? pointsFromMeta(r.meta) : undefined,
+      metrics: asObject(r.metrics),
+      meta: asObject(r.meta),
+      deviceBindingId: r.device_binding_id ?? null,
+      lineId: r.line_id ?? null,
+      productId: r.product_id ?? null,
+      recipeId: r.recipe_id ?? null,
+      runId: r.run_id ?? null,
+    }))
+  }
+}
+
+/** 向量点列从 meta.points 还原(非数组/越界防御) */
+function pointsFromMeta(meta: unknown): number[] | undefined {
+  const p = asObject(meta).points
+  if (!Array.isArray(p) || p.length === 0 || p.length > 4096) return undefined
+  return p.every(x => Number.isFinite(Number(x))) ? p.map(Number) : undefined
+}
+
+function asObject(v: unknown): Record<string, unknown> {
+  if (typeof v === 'string') {
+    try {
+      return JSON.parse(v) as Record<string, unknown>
+    }
+    catch {
+      return {}
+    }
+  }
+  return (v && typeof v === 'object' ? v : {}) as Record<string, unknown>
+}
+
+/** timestamptz → epoch ms(pg 返回 Date 对象;Date.parse(Date) 会经本地时区字符串丢毫秒) */
+function tsToMs(v: unknown): number {
+  if (v instanceof Date) return v.getTime()
+  if (typeof v === 'number') return v
+  return Date.parse(String(v))
 }

package/server/services/workshop/daq/storage/tsdb-port.ts

@@ -69,6 +69,52 @@ export interface TsdbPort {
   queryTagged(q: TsdbTagQuery): Promise<Map<string, TsdbPoint[]>>
   /** 各节点最近一个样本(列表页值列兜底) */
   latest(): Promise<Map<string, DaqSampleRow>>
+  // ===== 多形态帧(v2:向量/图像;像素 blob 在对象存储,这里只存元数据+派生指标)=====
+  writeFrames(rows: DaqFrameRow[]): Promise<void>
+  queryFrames(nodeId: string, opts: DaqFrameQueryOpts): Promise<DaqFrameRecord[]>
   /** 释放底层连接池/句柄(rebuild 换池时由工厂调用;可选) */
   close?(): Promise<void> | void
 }
+
+// ===== 帧(daq_frames;Timescale 元数据 + JSONB 向量/指标,图像只存对象键)=====
+
+export interface DaqFrameRow {
+  nodeId: string
+  tsMs: number
+  kind: 'vector' | 'image'
+  templateKey?: string | null
+  deviceBindingId?: string | null
+  lineId?: string | null
+  productId?: string | null
+  recipeId?: string | null
+  runId?: string | null
+  /** 向量点数(图像为 0) */
+  points: number
+  /** 图像元数据 { objectKey, thumbKey, mime, width, height };向量 {} */
+  meta: Record<string, unknown>
+  /** 派生指标(avg/max/brightness/zone_*…) */
+  metrics: Record<string, number>
+}
+
+export interface DaqFrameQueryOpts {
+  fromMs?: number
+  toMs?: number
+  kind?: 'vector' | 'image'
+  limit?: number
+}
+
+/** 帧查询记录(REST 返回形态;blob 按需经对象存储取) */
+export interface DaqFrameRecord {
+  at: number
+  kind: 'vector' | 'image'
+  /** 向量完整点列(图像缺省) */
+  points?: number[]
+  metrics: Record<string, number>
+  /** 图像元数据(objectKey/thumbKey/mime/width/height) */
+  meta: Record<string, unknown>
+  deviceBindingId?: string | null
+  lineId?: string | null
+  productId?: string | null
+  recipeId?: string | null
+  runId?: string | null
+}