Create live Phase-1 extraction jobs in PostgreSQL

Constraint: Continue Phase-1 industrialization without waiting on missing audio mounts, and keep every Ralph step documented and pushed
Rejected: Leave extraction scheduling as an implicit next step after registry bootstrap | It forces future sessions to reconstruct pending jobs by hand
Confidence: high
Scope-risk: narrow
Directive: Use feature_extraction_job as the canonical handoff between registry bootstrap and actual encoder extraction runs
Tested: /usr/local/miniconda3/bin/python scripts/bootstrap_phase1_extraction_jobs_live.py --dsn 'postgres://d2:d2pass@127.0.0.1:5432/d2' --schema acr_test --output data/pgvector_eval/music20/phase1_extraction_jobs_report.json; /usr/local/miniconda3/bin/python -m py_compile scripts/bootstrap_phase1_extraction_jobs_live.py; git diff --check -- acr-engine/scripts/bootstrap_phase1_extraction_jobs_live.py acr-engine/data/pgvector_eval/music20/phase1_extraction_jobs_report.json docs/model-feature-registry-bootstrap.md docs/postgres_db_schema_samples.md docs/session-handoff.md docs/CHANGELOG.md
Not-tested: Downstream worker that consumes these pending jobs to run real MERT/MuQ extraction

+{
+  "schema": "acr_test",
+  "dsn_redacted": "postgres://d2:***@127.0.0.1:5432/d2",
+  "jobs": [
+    {
+      "extraction_job_id": 1,
+      "feature_set_id": 2,
+      "model_name": "chromaprint",
+      "model_version": "v1",
+      "feature_name": "fingerprint_asset",
+      "window_sec": 5.0,
+      "hop_sec": 2.5,
+      "target_scope": "reference_set:phase1_hot_reference_v1",
+      "job_status": "pending",
+      "operation": "inserted"
+    },
+    {
+      "extraction_job_id": 2,
+      "feature_set_id": 3,
+      "model_name": "mert",
+      "model_version": "v1-95m",
+      "feature_name": "semantic_embedding",
+      "window_sec": 5.0,
+      "hop_sec": 2.5,
+      "target_scope": "reference_set:phase1_hot_reference_v1",
+      "job_status": "pending",
+      "operation": "inserted"
+    },
+    {
+      "extraction_job_id": 3,
+      "feature_set_id": 4,
+      "model_name": "mert",
+      "model_version": "v1-95m",
+      "feature_name": "semantic_embedding",
+      "window_sec": 10.0,
+      "hop_sec": 5.0,
+      "target_scope": "reference_set:phase1_hot_reference_v1",
+      "job_status": "pending",
+      "operation": "inserted"
+    },
+    {
+      "extraction_job_id": 4,
+      "feature_set_id": 5,
+      "model_name": "muq",
+      "model_version": "large-msd-iter",
+      "feature_name": "semantic_embedding",
+      "window_sec": 5.0,
+      "hop_sec": 2.5,
+      "target_scope": "reference_set:phase1_hot_reference_v1",
+      "job_status": "pending",
+      "operation": "inserted"
+    },
+    {
+      "extraction_job_id": 5,
+      "feature_set_id": 6,
+      "model_name": "ecapa",
+      "model_version": "acr-baseline-v1",
+      "feature_name": "semantic_embedding",
+      "window_sec": 5.0,
+      "hop_sec": 2.5,
+      "target_scope": "reference_set:phase1_hot_reference_v1",
+      "job_status": "pending",
+      "operation": "inserted"
+    }
+  ],
+  "counts": {
+    "feature_extraction_job": 5,
+    "pending_jobs": 5
+  }
+}
\ No newline at end of file

+#!/usr/bin/env /usr/local/miniconda3/bin/python
+from __future__ import annotations
+
+import argparse
+import json
+from pathlib import Path
+from typing import Any
+
+import psycopg
+
+ROOT = Path(__file__).resolve().parents[1]
+DEFAULT_OUTPUT = ROOT / 'data' / 'pgvector_eval' / 'music20' / 'phase1_extraction_jobs_report.json'
+
+JOB_SPECS = [
+    {
+        'model_name': 'chromaprint',
+        'model_version': 'v1',
+        'feature_name': 'fingerprint_asset',
+        'window_sec': 5.0,
+        'hop_sec': 2.5,
+        'target_scope': 'reference_set:phase1_hot_reference_v1',
+        'job_status': 'pending',
+        'shard_key': 'phase1/reference/chromaprint/v1',
+        'metadata_json': {'lane': 'exact', 'phase': 'phase1', 'priority': 'p0'},
+    },
+    {
+        'model_name': 'mert',
+        'model_version': 'v1-95m',
+        'feature_name': 'semantic_embedding',
+        'window_sec': 5.0,
+        'hop_sec': 2.5,
+        'target_scope': 'reference_set:phase1_hot_reference_v1',
+        'job_status': 'pending',
+        'shard_key': 'phase1/reference/mert/v1-95m/5s_2.5s',
+        'metadata_json': {'lane': 'semantic', 'role': 'primary_baseline', 'phase': 'phase1'},
+    },
+    {
+        'model_name': 'mert',
+        'model_version': 'v1-95m',
+        'feature_name': 'semantic_embedding',
+        'window_sec': 10.0,
+        'hop_sec': 5.0,
+        'target_scope': 'reference_set:phase1_hot_reference_v1',
+        'job_status': 'pending',
+        'shard_key': 'phase1/reference/mert/v1-95m/10s_5s',
+        'metadata_json': {'lane': 'semantic', 'role': 'long_context_validation', 'phase': 'phase1'},
+    },
+    {
+        'model_name': 'muq',
+        'model_version': 'large-msd-iter',
+        'feature_name': 'semantic_embedding',
+        'window_sec': 5.0,
+        'hop_sec': 2.5,
+        'target_scope': 'reference_set:phase1_hot_reference_v1',
+        'job_status': 'pending',
+        'shard_key': 'phase1/reference/muq/large-msd-iter/5s_2.5s',
+        'metadata_json': {'lane': 'semantic', 'role': 'challenger', 'phase': 'phase1'},
+    },
+    {
+        'model_name': 'ecapa',
+        'model_version': 'acr-baseline-v1',
+        'feature_name': 'semantic_embedding',
+        'window_sec': 5.0,
+        'hop_sec': 2.5,
+        'target_scope': 'reference_set:phase1_hot_reference_v1',
+        'job_status': 'pending',
+        'shard_key': 'phase1/reference/ecapa/acr-baseline-v1/5s_2.5s',
+        'metadata_json': {'lane': 'semantic', 'role': 'historical_baseline', 'phase': 'phase1'},
+    },
+]
+
+
+def resolve_feature_set_id(conn: psycopg.Connection, job: dict[str, Any]) -> int:
+    row = conn.execute(
+        """
+        SELECT fs.feature_set_id
+        FROM feature_set_registry fs
+        JOIN model_registry mr ON mr.model_id = fs.model_id
+        WHERE mr.model_name = %s
+          AND mr.model_version = %s
+          AND fs.feature_name = %s
+          AND coalesce(fs.window_sec, -1) = coalesce(%s, -1)
+          AND coalesce(fs.hop_sec, -1) = coalesce(%s, -1)
+        ORDER BY fs.feature_set_id
+        LIMIT 1;
+        """,
+        (
+            job['model_name'],
+            job['model_version'],
+            job['feature_name'],
+            job['window_sec'],
+            job['hop_sec'],
+        ),
+    ).fetchone()
+    if not row:
+        raise RuntimeError(
+            f"Feature set not found for {job['model_name']} {job['model_version']} {job['feature_name']} {job['window_sec']}/{job['hop_sec']}"
+        )
+    return int(row[0])
+
+
+def ensure_job(conn: psycopg.Connection, feature_set_id: int, job: dict[str, Any]) -> tuple[int, str]:
+    existing = conn.execute(
+        """
+        SELECT extraction_job_id
+        FROM feature_extraction_job
+        WHERE feature_set_id = %s
+          AND target_scope = %s
+          AND coalesce(shard_key, '') = coalesce(%s, '')
+        ORDER BY extraction_job_id
+        LIMIT 1;
+        """,
+        (feature_set_id, job['target_scope'], job['shard_key']),
+    ).fetchone()
+    if existing:
+        conn.execute(
+            """
+            UPDATE feature_extraction_job
+            SET job_status = %s,
+                metadata_json = %s::jsonb
+            WHERE extraction_job_id = %s;
+            """,
+            (job['job_status'], json.dumps(job['metadata_json']), existing[0]),
+        )
+        return int(existing[0]), 'reused'
+
+    row = conn.execute(
+        """
+        INSERT INTO feature_extraction_job (
+            feature_set_id, target_scope, job_status, shard_key,
+            input_count, output_count, started_at, finished_at,
+            log_uri, metadata_json
+        ) VALUES (
+            %s, %s, %s, %s,
+            NULL, NULL, NULL, NULL,
+            NULL, %s::jsonb
+        )
+        RETURNING extraction_job_id;
+        """,
+        (feature_set_id, job['target_scope'], job['job_status'], job['shard_key'], json.dumps(job['metadata_json'])),
+    ).fetchone()
+    return int(row[0]), 'inserted'
+
+
+def main() -> None:
+    ap = argparse.ArgumentParser()
+    ap.add_argument('--dsn', required=True)
+    ap.add_argument('--schema', default='acr_test')
+    ap.add_argument('--output', default=str(DEFAULT_OUTPUT))
+    args = ap.parse_args()
+
+    summary: dict[str, Any] = {
+        'schema': args.schema,
+        'dsn_redacted': 'postgres://d2:***@127.0.0.1:5432/d2',
+        'jobs': [],
+    }
+    with psycopg.connect(args.dsn, autocommit=True) as conn:
+        conn.execute(f'SET search_path TO {args.schema}, public;')
+        for job in JOB_SPECS:
+            feature_set_id = resolve_feature_set_id(conn, job)
+            extraction_job_id, operation = ensure_job(conn, feature_set_id, job)
+            summary['jobs'].append({
+                'extraction_job_id': extraction_job_id,
+                'feature_set_id': feature_set_id,
+                'model_name': job['model_name'],
+                'model_version': job['model_version'],
+                'feature_name': job['feature_name'],
+                'window_sec': job['window_sec'],
+                'hop_sec': job['hop_sec'],
+                'target_scope': job['target_scope'],
+                'job_status': job['job_status'],
+                'operation': operation,
+            })
+        summary['counts'] = {
+            'feature_extraction_job': int(conn.execute('SELECT count(*) FROM feature_extraction_job;').fetchone()[0]),
+            'pending_jobs': int(conn.execute("SELECT count(*) FROM feature_extraction_job WHERE job_status = 'pending';").fetchone()[0]),
+        }
+
+    out = Path(args.output)
+    out.parent.mkdir(parents=True, exist_ok=True)
+    out.write_text(json.dumps(summary, ensure_ascii=False, indent=2), encoding='utf-8')
+    print(json.dumps(summary, ensure_ascii=False, indent=2))
+
+
+if __name__ == '__main__':
+    main()

 ## 2026-06-04
 
+- 新增 `acr-engine/scripts/bootstrap_phase1_extraction_jobs_live.py` 与 `acr-engine/data/pgvector_eval/music20/phase1_extraction_jobs_report.json`，把 Phase-1 的 `feature_extraction_job` 初始化做成可直接连 PostgreSQL 的 live 脚本，并已在 `acr_test` schema 真实创建 5 条 pending jobs。
 - 补充 `phase1_registry_bootstrap_idempotency_report.json` 与文档说明，验证 `bootstrap_phase1_model_registry_live.py` 在 `acr_test` schema 上连续执行两次后表计数保持稳定，证明 Phase-1 registry bootstrap 具备可重复执行的幂等性。
 - 新增 `acr-engine/scripts/bootstrap_phase1_model_registry_live.py` 与 `acr-engine/data/pgvector_eval/music20/phase1_registry_bootstrap_report.json`，把 Phase-1 的 `chromaprint / mert / muq / ecapa` 与对应 `feature_set_registry / reference_set_registry` 初始化做成可直接连 PostgreSQL 的 live bootstrap 脚本，并已在 `acr_test` schema 验证通过。
 - 补充文档阻塞事实：当前容器里缺少 `/workspace/downloads`，因此本轮无法直接从业务样本目录继续生成 `type_8 / type_16` 的 live PostgreSQL query JSONL；已把该环境前提写入 handoff 与 PostgreSQL 样例文档。

@@ -300,3 +300,47 @@ cd /workspace/acr-engine
 结论：
 
 > 当前 bootstrap 脚本可重复执行，不会把 Phase-1 registry 数据重复灌爆。
+
+---
+
+## 9. Phase-1 extraction job bootstrap
+
+当 `model_registry / feature_set_registry / reference_set_registry` 都已经存在后，下一步不是立刻手工跑抽特征，而是先把 **待执行 job** 写到 `feature_extraction_job`。
+
+本仓库现在已经提供：
+
+- `acr-engine/scripts/bootstrap_phase1_extraction_jobs_live.py`
+
+用途：
+- 根据已存在的 `feature_set_registry`
+- 为 `phase1_hot_reference_v1` 生成待执行 extraction jobs
+- 把 Phase-1 的 exact / semantic lanes 统一放进 PostgreSQL job 表
+
+### 9.1 执行命令
+
+```bash
+cd /workspace/acr-engine
+/usr/local/miniconda3/bin/python scripts/bootstrap_phase1_extraction_jobs_live.py \
+  --dsn 'postgres://d2:d2pass@127.0.0.1:5432/d2' \
+  --schema acr_test \
+  --output data/pgvector_eval/music20/phase1_extraction_jobs_report.json
+```
+
+### 9.2 当前已验证结果（acr_test）
+
+本轮已真实创建 5 条待执行 job：
+
+| lane | model | feature | target_scope | status |
+|---|---|---|---|---|
+| exact | `chromaprint` | `fingerprint_asset` | `reference_set:phase1_hot_reference_v1` | `pending` |
+| semantic | `mert` | `semantic_embedding` 5s/2.5s | `reference_set:phase1_hot_reference_v1` | `pending` |
+| semantic | `mert` | `semantic_embedding` 10s/5s | `reference_set:phase1_hot_reference_v1` | `pending` |
+| semantic | `muq` | `semantic_embedding` 5s/2.5s | `reference_set:phase1_hot_reference_v1` | `pending` |
+| semantic | `ecapa` | `semantic_embedding` 5s/2.5s | `reference_set:phase1_hot_reference_v1` | `pending` |
+
+对应 live 报告：
+- `acr-engine/data/pgvector_eval/music20/phase1_extraction_jobs_report.json`
+
+这意味着：
+
+> 现在 PostgreSQL 里已经不只是“模型定义”和“特征定义”，而是连 **下一步该跑哪些抽特征任务** 都已经具备结构化入口了。

@@ -67,6 +67,7 @@
 | FAISS 对照报告 | `acr-engine/data/pgvector_eval/music20/songid_eval_report_fresh.json` |
 | registry bootstrap 报告 | `acr-engine/data/pgvector_eval/music20/phase1_registry_bootstrap_report.json` |
 | registry bootstrap 幂等性报告 | `acr-engine/data/pgvector_eval/music20/phase1_registry_bootstrap_idempotency_report.json` |
+| extraction job bootstrap 报告 | `acr-engine/data/pgvector_eval/music20/phase1_extraction_jobs_report.json` |
 | 历史对照报告 | `acr-engine/data/pgvector_eval/music20/songid_eval_report.json` |
 
 ---
@@ -399,6 +400,22 @@ flowchart LR
 对应 live 报告：
 - `acr-engine/data/pgvector_eval/music20/phase1_registry_bootstrap_report.json`
 
+### 本轮继续新增：Phase-1 extraction jobs 已可 live bootstrap
+
+在 registry bootstrap 之后，本轮又新增：
+
+- `acr-engine/scripts/bootstrap_phase1_extraction_jobs_live.py`
+
+它已经在 `acr_test` schema 上真实创建了 5 条 `feature_extraction_job`：
+- `chromaprint`
+- `mert 5s/2.5s`
+- `mert 10s/5s`
+- `muq 5s/2.5s`
+- `ecapa 5s/2.5s`
+
+对应 live 报告：
+- `acr-engine/data/pgvector_eval/music20/phase1_extraction_jobs_report.json`
+
 ### 路线 1：继续做 PostgreSQL 工程化
 
 1. 把 `live_pgvector_music20_eval.py` 泛化成：

@@ -183,6 +183,7 @@ sed -n '1,320p' acr-engine/sql/acr_pg_schema_v2.sql
 - 机械校验已补齐：`live_pgvector_music20_eval.py` 的 `py_compile` 通过，相关变更 `diff --check` 通过
 - PostgreSQL `acr_test` schema 上已真实写入 Phase-1 registry bootstrap：`chromaprint / mert / muq / ecapa` + 5 组 feature set + `phase1_hot_reference_v1`
 - Phase-1 registry bootstrap 已有幂等性证据：同 schema 连续执行两次后，`model_registry=5 / feature_set_registry=6 / reference_set_registry=2` 保持不变
+- PostgreSQL `acr_test` schema 上已真实创建 5 条 `feature_extraction_job`，后续 MERT / MuQ 接入可直接从 pending jobs 启动
 
 ### 未验证 / 仍是缺口
 - **未实际跑 MERT / MuQ encoder-only 特征抽取**