@ngocsangairvds/vsaf 3.2.14 → 3.2.16

package/tools/vds-scripts/audit_orchestrator/alembic/versions/0001_initial_audit_state_schema.py

@@ -1,1260 +0,0 @@
-"""Initial audit_state schema (squash of phase47..phase170 bootstrap)
-
-Revision ID: 0001
-Revises:
-Create Date: 2026-04-25
-
-P4 / TSK-4.1.4: subsumes ALL prior schema sources for audit_state into a
-single squash migration:
-
-  (a) BOOTSTRAP_SQL constants from
-      vds_audit_orchestrator.state.store
-        - _BOOTSTRAP_TABLES_SQL  (Phase A: schema + tables + constraints)
-        - _BOOTSTRAP_INDEXES_SQL (Phase B: indexes)
-        - _BOOTSTRAP_FUNCTIONS_SQL (Phase C: functions + view)
-
-  (b) state/migrations/*.sql snapshots that carried DDL not in BOOTSTRAP:
-        - phase157_dispatch_tables.sql -> ALTER TABLE dispatch_lanes
-              ADD COLUMN IF NOT EXISTS claimed_at TIMESTAMPTZ
-              (NOT in _BOOTSTRAP_TABLES_SQL — additive Phase 168 column;
-              required by _DISPATCH_SCHEMA_CONTRACT runtime contract)
-        - phase117/118, phase154, phase157 (rest), phase169,
-          state_repair_hardening — all already in BOOTSTRAP_SQL
-          (verified by diff; retained as documentation in legacy/migrations).
-
-Idempotence: every CREATE / ALTER / INSERT below uses IF NOT EXISTS,
-ON CONFLICT, or DROP IF EXISTS so this migration is safe to replay against
-an existing live database via:
-
-    alembic stamp 0001     # mark already-bootstrapped DBs as at head
-    alembic upgrade head   # no-op afterwards
-
-Pre-existing operator DBs were bootstrapped by store.init_schema() which
-ran the same DDL. Stamping is the recommended path; running upgrade
-without stamping is also safe due to the idempotence guards.
-
-Downgrade:
-    Pass-through. Schema retirement is a manual operator decision (data
-    preservation > data destruction), NOT an alembic concern. We do NOT
-    DROP SCHEMA audit_state CASCADE.
-"""
-
-from __future__ import annotations
-
-from collections.abc import Sequence
-
-import sqlalchemy as sa
-from alembic import op
-
-# revision identifiers, used by Alembic.
-revision: str = "0001"
-down_revision: str | None = None
-branch_labels: str | Sequence[str] | None = None
-depends_on: str | Sequence[str] | None = None
-
-
-# ─── Phase A: Schema + tables + constraints (transactional) ──────────────────
-# Mirrors _BOOTSTRAP_TABLES_SQL in state/store.py.
-# Includes the additive Phase 168 ALTER (dispatch_lanes.claimed_at).
-
-_BOOTSTRAP_EXTENSIONS_SQL = """
-CREATE EXTENSION IF NOT EXISTS vector;
-CREATE EXTENSION IF NOT EXISTS unaccent;
-"""
-
-_BOOTSTRAP_TABLES_SQL = """
-CREATE SCHEMA IF NOT EXISTS audit_state;
-
--- Phase 47/50: schema_version anchor
-CREATE TABLE IF NOT EXISTS audit_state.schema_version (
-    id INTEGER PRIMARY KEY DEFAULT 1 CHECK (id = 1),
-    version INTEGER NOT NULL CHECK (version > 0),
-    phase TEXT NOT NULL DEFAULT 'phase50',
-    applied_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
-);
-
--- Hierarchy: departments → projects → repos
-CREATE TABLE IF NOT EXISTS audit_state.departments (
-    id BIGSERIAL PRIMARY KEY,
-    department_key TEXT NOT NULL UNIQUE,
-    display_name TEXT NOT NULL,
-    is_quarantined BOOLEAN NOT NULL DEFAULT FALSE,
-    quarantine_reason TEXT,
-    metadata JSONB NOT NULL DEFAULT '{}'::jsonb,
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
-);
-
-CREATE TABLE IF NOT EXISTS audit_state.bitbucket_projects (
-    id BIGSERIAL PRIMARY KEY,
-    project_key TEXT NOT NULL UNIQUE,
-    display_name TEXT NOT NULL,
-    is_quarantined BOOLEAN NOT NULL DEFAULT FALSE,
-    quarantine_reason TEXT,
-    metadata JSONB NOT NULL DEFAULT '{}'::jsonb,
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
-);
-
-CREATE TABLE IF NOT EXISTS audit_state.projects (
-    id BIGSERIAL PRIMARY KEY,
-    project_key TEXT NOT NULL UNIQUE,
-    display_name TEXT NOT NULL,
-    department_id BIGINT REFERENCES audit_state.departments(id) ON DELETE SET NULL,
-    checklist_page_id TEXT,
-    metadata JSONB NOT NULL DEFAULT '{}'::jsonb,
-    state_version INTEGER NOT NULL DEFAULT 1 CHECK (state_version > 0),
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
-);
-
-CREATE TABLE IF NOT EXISTS audit_state.repos (
-    id BIGSERIAL PRIMARY KEY,
-    project_id BIGINT NOT NULL REFERENCES audit_state.projects(id) ON DELETE CASCADE,
-    repo_key TEXT NOT NULL UNIQUE,
-    repo_url TEXT NOT NULL,
-    default_branch TEXT NOT NULL DEFAULT 'main',
-    head_commit_sha TEXT NOT NULL DEFAULT '',
-    repo_type TEXT NOT NULL DEFAULT 'primary' CHECK (repo_type IN ('primary', 'ops')),
-    bitbucket_project_id BIGINT REFERENCES audit_state.bitbucket_projects(id) ON DELETE SET NULL,
-    analysis_excluded BOOLEAN NOT NULL DEFAULT FALSE,
-    analysis_exclusion_reason TEXT NOT NULL DEFAULT '',
-    analysis_exclusion_updated_at TIMESTAMPTZ,
-    metadata JSONB NOT NULL DEFAULT '{}'::jsonb,
-    state_version INTEGER NOT NULL DEFAULT 1 CHECK (state_version > 0),
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
-);
-
-CREATE TABLE IF NOT EXISTS audit_state.artifacts (
-    id BIGSERIAL PRIMARY KEY,
-    repo_id BIGINT NOT NULL REFERENCES audit_state.repos(id) ON DELETE CASCADE,
-    artifact_key TEXT NOT NULL UNIQUE,
-    artifact_type TEXT NOT NULL,
-    status TEXT NOT NULL DEFAULT 'pending',
-    payload JSONB NOT NULL DEFAULT '{}'::jsonb,
-    claimed_by TEXT,
-    claimed_at TIMESTAMPTZ,
-    claim_expires_at TIMESTAMPTZ,
-    state_version INTEGER NOT NULL DEFAULT 1 CHECK (state_version > 0),
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
-);
-
-CREATE TABLE IF NOT EXISTS audit_state.runs (
-    id BIGSERIAL PRIMARY KEY,
-    run_id TEXT NOT NULL UNIQUE,
-    project_key TEXT NOT NULL,
-    repo_key TEXT,
-    run_type TEXT NOT NULL DEFAULT 'workflow',
-    status TEXT NOT NULL DEFAULT 'pending',
-    payload JSONB NOT NULL DEFAULT '{}'::jsonb,
-    started_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    finished_at TIMESTAMPTZ,
-    state_version INTEGER NOT NULL DEFAULT 1 CHECK (state_version > 0),
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
-);
-
-CREATE TABLE IF NOT EXISTS audit_state.profile_performance (
-    id BIGSERIAL PRIMARY KEY,
-    profile_name TEXT NOT NULL,
-    project_key TEXT NOT NULL,
-    repo_key TEXT NOT NULL,
-    run_id TEXT NOT NULL,
-    overall_score DOUBLE PRECISION,
-    pass_rate DOUBLE PRECISION,
-    error_rate DOUBLE PRECISION,
-    timeout_rate DOUBLE PRECISION,
-    avg_row_duration_ms DOUBLE PRECISION,
-    docs_hit_rate DOUBLE PRECISION,
-    code_hit_rate DOUBLE PRECISION,
-    row_count INTEGER NOT NULL DEFAULT 0 CHECK (row_count >= 0),
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    CONSTRAINT uq_audit_state_profile_performance_run UNIQUE (profile_name, project_key, repo_key, run_id)
-);
-
-CREATE TABLE IF NOT EXISTS audit_state.run_events (
-    id BIGSERIAL PRIMARY KEY,
-    run_pk BIGINT NOT NULL REFERENCES audit_state.runs(id) ON DELETE CASCADE,
-    event_type TEXT NOT NULL CHECK (event_type IN ('run_started', 'analysis_completed', 'publish_completed', 'run_failed', 'run_resumed')),
-    status TEXT NOT NULL DEFAULT 'info',
-    payload JSONB NOT NULL DEFAULT '{}'::jsonb,
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
-);
-
--- Phase 50: checklist run/result/artifact tables
-CREATE TABLE IF NOT EXISTS audit_state.checklist_runs (
-    id BIGSERIAL PRIMARY KEY,
-    run_pk BIGINT NOT NULL UNIQUE REFERENCES audit_state.runs(id) ON DELETE CASCADE,
-    thread_id TEXT NOT NULL,
-    run_id TEXT NOT NULL UNIQUE REFERENCES audit_state.runs(run_id) ON DELETE CASCADE,
-    project_key TEXT NOT NULL,
-    repo_key TEXT,
-    checklist_version TEXT NOT NULL DEFAULT '',
-    checklist_page_id TEXT,
-    projection_profile_version TEXT,
-    provider TEXT,
-    model TEXT,
-    runtime_profile TEXT,
-    retrieval_mode TEXT NOT NULL DEFAULT 'hybrid',
-    status TEXT NOT NULL DEFAULT 'pending',
-    payload JSONB NOT NULL DEFAULT '{}'::jsonb,
-    started_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    finished_at TIMESTAMPTZ,
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    CONSTRAINT ck_audit_state_checklist_runs_thread_id_not_blank CHECK (length(btrim(thread_id)) > 0),
-    CONSTRAINT ck_audit_state_checklist_runs_checklist_version_not_blank CHECK (length(btrim(checklist_version)) > 0),
-    CONSTRAINT ck_audit_state_checklist_runs_retrieval_mode_not_blank CHECK (length(btrim(retrieval_mode)) > 0),
-    CONSTRAINT ck_audit_state_checklist_runs_status_not_blank CHECK (length(btrim(status)) > 0)
-);
-
-CREATE TABLE IF NOT EXISTS audit_state.checklist_row_results (
-    id BIGSERIAL PRIMARY KEY,
-    checklist_run_id BIGINT NOT NULL REFERENCES audit_state.checklist_runs(id) ON DELETE CASCADE,
-    row_key TEXT NOT NULL,
-    row_index INTEGER NOT NULL CHECK (row_index >= 0),
-    check_id TEXT,
-    status TEXT NOT NULL,
-    score DOUBLE PRECISION,
-    finding TEXT,
-    reason TEXT,
-    evidence_refs JSONB NOT NULL DEFAULT '[]'::jsonb,
-    provenance JSONB NOT NULL DEFAULT '{}'::jsonb,
-    retry_count INTEGER NOT NULL DEFAULT 0 CHECK (retry_count >= 0),
-    retry_metadata JSONB NOT NULL DEFAULT '{}'::jsonb,
-    payload JSONB NOT NULL DEFAULT '{}'::jsonb,
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    CONSTRAINT uq_audit_state_checklist_row_results_run_row UNIQUE(checklist_run_id, row_key),
-    CONSTRAINT ck_audit_state_checklist_row_results_row_key_not_blank CHECK (length(btrim(row_key)) > 0),
-    CONSTRAINT ck_audit_state_checklist_row_results_status_not_blank CHECK (length(btrim(status)) > 0)
-);
-
-CREATE TABLE IF NOT EXISTS audit_state.checklist_artifacts (
-    id BIGSERIAL PRIMARY KEY,
-    checklist_run_id BIGINT NOT NULL REFERENCES audit_state.checklist_runs(id) ON DELETE CASCADE,
-    artifact_type TEXT NOT NULL,
-    artifact_path TEXT NOT NULL,
-    artifact_hash TEXT NOT NULL,
-    artifact_version TEXT,
-    metadata JSONB NOT NULL DEFAULT '{}'::jsonb,
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    CONSTRAINT uq_audit_state_checklist_artifacts_run_type UNIQUE(checklist_run_id, artifact_type),
-    CONSTRAINT ck_audit_state_checklist_artifacts_type_not_blank CHECK (length(btrim(artifact_type)) > 0),
-    CONSTRAINT ck_audit_state_checklist_artifacts_path_not_blank CHECK (length(btrim(artifact_path)) > 0),
-    CONSTRAINT ck_audit_state_checklist_artifacts_hash_not_blank CHECK (length(btrim(artifact_hash)) > 0)
-);
-
--- Evidence + retrieval snapshots
-CREATE TABLE IF NOT EXISTS audit_state.evidence_snapshots (
-    id BIGSERIAL PRIMARY KEY,
-    snapshot_key TEXT NOT NULL UNIQUE,
-    project_key TEXT NOT NULL,
-    repo_key TEXT,
-    content_hash TEXT NOT NULL,
-    payload JSONB NOT NULL DEFAULT '{}'::jsonb,
-    reuse_count INTEGER NOT NULL DEFAULT 0 CHECK (reuse_count >= 0),
-    first_seen_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    last_reused_at TIMESTAMPTZ,
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
-);
-
-CREATE TABLE IF NOT EXISTS audit_state.retrieval_snapshots (
-    id BIGSERIAL PRIMARY KEY,
-    snapshot_key TEXT NOT NULL UNIQUE,
-    project_key TEXT NOT NULL,
-    repo_key TEXT,
-    embedding_model_name TEXT NOT NULL,
-    embedding_model_digest TEXT NOT NULL DEFAULT '',
-    embedding_dimensions INTEGER NOT NULL CHECK (embedding_dimensions > 0),
-    index_type TEXT NOT NULL DEFAULT 'hnsw',
-    lexical_only BOOLEAN NOT NULL DEFAULT TRUE,
-    embedding_model_changed BOOLEAN NOT NULL DEFAULT FALSE,
-    force_reason TEXT,
-    content_hash TEXT NOT NULL DEFAULT '',
-    payload JSONB NOT NULL DEFAULT '{}'::jsonb,
-    chunk_count INTEGER NOT NULL DEFAULT 0 CHECK (chunk_count >= 0),
-    changed_chunk_count INTEGER NOT NULL DEFAULT 0 CHECK (changed_chunk_count >= 0),
-    state_version INTEGER NOT NULL DEFAULT 1 CHECK (state_version > 0),
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
-);
-
--- Project / repo links
-CREATE TABLE IF NOT EXISTS audit_state.project_links (
-    id BIGSERIAL PRIMARY KEY,
-    project_id BIGINT NOT NULL REFERENCES audit_state.projects(id) ON DELETE CASCADE,
-    link_type TEXT NOT NULL,
-    link_role TEXT NOT NULL DEFAULT '',
-    normalized_url TEXT NOT NULL,
-    raw_url TEXT NOT NULL,
-    source_page_id TEXT NOT NULL DEFAULT '',
-    source_page_url TEXT NOT NULL DEFAULT '',
-    source_context TEXT NOT NULL DEFAULT '',
-    last_seen_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    metadata JSONB NOT NULL DEFAULT '{}'::jsonb,
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    CONSTRAINT uq_audit_state_project_links UNIQUE(
-        project_id,
-        link_type,
-        link_role,
-        normalized_url,
-        source_page_id,
-        source_page_url,
-        source_context
-    )
-);
-
-CREATE TABLE IF NOT EXISTS audit_state.repo_links (
-    id BIGSERIAL PRIMARY KEY,
-    repo_id BIGINT NOT NULL REFERENCES audit_state.repos(id) ON DELETE CASCADE,
-    link_type TEXT NOT NULL,
-    link_role TEXT NOT NULL DEFAULT '',
-    normalized_url TEXT NOT NULL,
-    raw_url TEXT NOT NULL,
-    source_page_id TEXT NOT NULL DEFAULT '',
-    last_seen_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    metadata JSONB NOT NULL DEFAULT '{}'::jsonb,
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    CONSTRAINT uq_audit_state_repo_links UNIQUE(
-        repo_id,
-        link_type,
-        link_role,
-        normalized_url,
-        source_page_id
-    )
-);
-
--- Project / repo preparation
-CREATE TABLE IF NOT EXISTS audit_state.project_preparation (
-    id BIGSERIAL PRIMARY KEY,
-    project_id BIGINT NOT NULL REFERENCES audit_state.projects(id) ON DELETE CASCADE,
-    status TEXT NOT NULL DEFAULT 'pending',
-    source TEXT,
-    payload_kind TEXT,
-    project_storage_key TEXT,
-    page_id TEXT,
-    project_status TEXT,
-    fetch_policy TEXT,
-    sync_metadata JSONB NOT NULL DEFAULT '{}'::jsonb,
-    prepared_at TIMESTAMPTZ,
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    CONSTRAINT uq_audit_state_project_preparation_project_id UNIQUE(project_id)
-);
-
-CREATE TABLE IF NOT EXISTS audit_state.repo_preparation (
-    id BIGSERIAL PRIMARY KEY,
-    repo_id BIGINT NOT NULL REFERENCES audit_state.repos(id) ON DELETE CASCADE,
-    status TEXT NOT NULL DEFAULT 'pending',
-    source TEXT,
-    payload_kind TEXT,
-    project_storage_key TEXT,
-    repo_storage_key TEXT,
-    repo_slug TEXT,
-    repo_url TEXT,
-    local_code_path TEXT,
-    sync_status TEXT,
-    sync_decision TEXT,
-    readiness TEXT,
-    prepared_at TIMESTAMPTZ,
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    CONSTRAINT uq_audit_state_repo_preparation_repo_id UNIQUE(repo_id)
-);
-
--- Phase 47: corpus tables (documents, chunks, embeddings)
-CREATE TABLE IF NOT EXISTS audit_state.corpus_documents (
-    id BIGSERIAL PRIMARY KEY,
-    snapshot_id BIGINT NOT NULL REFERENCES audit_state.retrieval_snapshots(id) ON DELETE CASCADE,
-    project_id BIGINT NOT NULL REFERENCES audit_state.projects(id) ON DELETE CASCADE,
-    project_storage_key TEXT,
-    repo_id BIGINT REFERENCES audit_state.repos(id) ON DELETE CASCADE,
-    document_id TEXT NOT NULL,
-    source_type TEXT NOT NULL DEFAULT 'confluence_page',
-    source_uri TEXT NOT NULL DEFAULT '',
-    doc_class TEXT,
-    doc_type_en TEXT,
-    version_tag TEXT NOT NULL DEFAULT '',
-    page_id TEXT,
-    content_hash TEXT NOT NULL,
-    title TEXT NULL,
-    page_title TEXT NOT NULL DEFAULT '',
-    space_key TEXT NOT NULL DEFAULT '',
-    metadata JSONB NOT NULL DEFAULT '{}'::jsonb,
-    parent_document_id BIGINT REFERENCES audit_state.corpus_documents(id) ON DELETE CASCADE,
-    state_version INTEGER NOT NULL DEFAULT 1 CHECK (state_version > 0),
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    CONSTRAINT ck_audit_state_corpus_documents_source_type CHECK (
-        source_type IN ('confluence_page', 'confluence_attachment', 'code_file', 'ops_repo_file', 'code_evidence_pack', 'bitbucket_readme')
-    ),
-    CONSTRAINT ck_audit_state_corpus_documents_confluence_page_id_numeric CHECK (
-        source_type <> 'confluence_page'
-        OR COALESCE(page_id, '') ~ '^[0-9]+$'
-    ),
-    CONSTRAINT ck_audit_state_corpus_documents_confluence_attachment_page_id_namespace CHECK (
-        source_type <> 'confluence_attachment'
-        OR COALESCE(page_id, '') ~ '^attachment:.+'
-    ),
-    CONSTRAINT ck_audit_state_corpus_documents_confluence_metadata_not_blank CHECK (
-        source_type NOT IN ('confluence_page', 'confluence_attachment')
-        OR (
-            length(btrim(COALESCE(page_title, ''))) > 0
-            AND length(btrim(COALESCE(space_key, ''))) > 0
-        )
-    ),
-    CONSTRAINT uq_audit_state_corpus_documents_snapshot_document UNIQUE(snapshot_id, document_id)
-);
-
-CREATE TABLE IF NOT EXISTS audit_state.corpus_chunks (
-    id BIGSERIAL PRIMARY KEY,
-    document_pk BIGINT NOT NULL REFERENCES audit_state.corpus_documents(id) ON DELETE CASCADE,
-    chunk_id TEXT NOT NULL,
-    chunk_order INTEGER,
-    start_line INTEGER NULL,
-    end_line INTEGER NULL,
-    page_id TEXT,
-    content_hash TEXT NOT NULL,
-    content_text TEXT NOT NULL,
-    metadata JSONB NOT NULL DEFAULT '{}'::jsonb,
-    search_vector tsvector GENERATED ALWAYS AS (
-        to_tsvector('simple', COALESCE(content_text, ''))
-    ) STORED,
-    state_version INTEGER NOT NULL DEFAULT 1 CHECK (state_version > 0),
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    CONSTRAINT ck_audit_state_corpus_chunks_chunk_order_positive CHECK (chunk_order >= 0),
-    CONSTRAINT ck_audit_state_corpus_chunks_line_range CHECK (end_line IS NULL OR start_line IS NULL OR end_line >= start_line),
-    CONSTRAINT ck_audit_state_corpus_chunks_content_text_length CHECK (length(content_text) <= 500000),
-    CONSTRAINT uq_audit_state_corpus_chunks_document_chunk UNIQUE(document_pk, chunk_id)
-);
-
-CREATE TABLE IF NOT EXISTS audit_state.corpus_embeddings (
-    id BIGSERIAL PRIMARY KEY,
-    snapshot_id BIGINT NOT NULL REFERENCES audit_state.retrieval_snapshots(id) ON DELETE CASCADE,
-    chunk_pk BIGINT NOT NULL REFERENCES audit_state.corpus_chunks(id) ON DELETE CASCADE,
-    embedding_model_name TEXT NOT NULL,
-    embedding_model_digest TEXT NOT NULL DEFAULT '',
-    embedding_dimensions INTEGER NOT NULL,
-    embedding_vector vector,
-    state_version INTEGER NOT NULL DEFAULT 1 CHECK (state_version > 0),
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    CONSTRAINT ck_audit_state_corpus_embeddings_dimensions_positive CHECK (embedding_dimensions > 0),
-    CONSTRAINT ck_audit_state_corpus_embeddings_vector_dimensions_match CHECK (
-        embedding_vector IS NULL OR vector_dims(embedding_vector) = embedding_dimensions
-    ),
-    CONSTRAINT uq_audit_state_corpus_embeddings_snapshot_chunk_model UNIQUE(snapshot_id, chunk_pk, embedding_model_digest)
-);
-
-CREATE TABLE IF NOT EXISTS audit_state.project_bitbucket_links (
-    id BIGSERIAL PRIMARY KEY,
-    project_id BIGINT NOT NULL REFERENCES audit_state.projects(id) ON DELETE CASCADE,
-    bitbucket_project_id BIGINT NOT NULL REFERENCES audit_state.bitbucket_projects(id) ON DELETE CASCADE,
-    source TEXT NOT NULL DEFAULT 'manual',
-    repo_filter TEXT,
-    metadata JSONB NOT NULL DEFAULT '{}'::jsonb,
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    UNIQUE(project_id, bitbucket_project_id)
-);
-
--- Phase 114: requirement classifications
-CREATE TABLE IF NOT EXISTS audit_state.requirement_classifications (
-    id BIGSERIAL PRIMARY KEY,
-    row_key TEXT NOT NULL,
-    content_fingerprint TEXT NOT NULL,
-    classification_fingerprint TEXT NOT NULL,
-    analysis_mode TEXT NOT NULL,
-    evidence_targets JSONB NOT NULL DEFAULT '[]'::jsonb,
-    project_scope_required BOOLEAN NOT NULL DEFAULT FALSE,
-    execution_scope_class TEXT NOT NULL DEFAULT 'repo_local',
-    finalization_policy TEXT NOT NULL DEFAULT 'default',
-    classification_source TEXT NOT NULL DEFAULT 'heuristic',
-    heuristic_confidence NUMERIC(5, 4) NOT NULL DEFAULT 0.5,
-    nlp_exemplar_similarity NUMERIC(5, 4) NOT NULL DEFAULT 0.0,
-    project_storage_key TEXT NOT NULL,
-    checklist_page_id TEXT NOT NULL,
-    run_id TEXT NOT NULL,
-    metadata JSONB NOT NULL DEFAULT '{}'::jsonb,
-    state_version INTEGER NOT NULL DEFAULT 1 CHECK (state_version > 0),
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    CONSTRAINT ck_audit_state_requirement_classifications_heuristic_confidence
-        CHECK (heuristic_confidence >= 0.0 AND heuristic_confidence <= 1.0),
-    CONSTRAINT ck_audit_state_requirement_classifications_nlp_similarity
-        CHECK (nlp_exemplar_similarity >= 0.0 AND nlp_exemplar_similarity <= 1.0),
-    CONSTRAINT ck_audit_state_requirement_classifications_execution_scope_class
-        CHECK (execution_scope_class IN ('repo_local', 'project_doc_authoritative', 'cross_repo_code_required')),
-    CONSTRAINT ck_audit_state_requirement_classifications_classification_source
-        CHECK (classification_source IN ('heuristic', 'nlp_enrichment', 'llm_agent', 'dspy_classifier', 'cached'))
-);
-
--- Phase 117/118: shared state additive tables
-CREATE TABLE IF NOT EXISTS audit_state.project_evidence_discoveries (
-    id TEXT PRIMARY KEY,
-    context_id TEXT NOT NULL,
-    project_key TEXT NOT NULL,
-    source_repo TEXT NOT NULL,
-    discovery_type TEXT NOT NULL,
-    ref TEXT NOT NULL,
-    excerpt TEXT NOT NULL,
-    confidence NUMERIC(5, 4) NOT NULL DEFAULT 0.0,
-    requirement_type_tag TEXT,
-    content_hash TEXT NOT NULL,
-    metadata JSONB NOT NULL DEFAULT '{}'::jsonb,
-    state_version INTEGER NOT NULL DEFAULT 1 CHECK (state_version > 0),
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    CONSTRAINT ck_audit_state_project_evidence_discoveries_confidence
-        CHECK (confidence >= 0.0 AND confidence <= 1.0),
-    CONSTRAINT uq_audit_state_project_evidence_discoveries_context_hash
-        UNIQUE(context_id, source_repo, content_hash)
-);
-
-CREATE TABLE IF NOT EXISTS audit_state.project_knowledge (
-    id TEXT PRIMARY KEY,
-    project_key TEXT NOT NULL,
-    repo_key TEXT,
-    knowledge_type TEXT NOT NULL,
-    content TEXT NOT NULL,
-    confidence NUMERIC(5, 4) NOT NULL DEFAULT 0.0,
-    source_run_id TEXT NOT NULL,
-    requirement_type_tag TEXT,
-    execution_scope_class TEXT,
-    source_context_id TEXT,
-    metadata JSONB NOT NULL DEFAULT '{}'::jsonb,
-    expires_at TIMESTAMPTZ,
-    retired_at TIMESTAMPTZ,
-    state_version INTEGER NOT NULL DEFAULT 1 CHECK (state_version > 0),
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    CONSTRAINT ck_audit_state_project_knowledge_confidence
-        CHECK (confidence >= 0.0 AND confidence <= 1.0),
-    CONSTRAINT ck_audit_state_project_knowledge_content_not_blank
-        CHECK (length(btrim(content)) > 0)
-);
-
--- Phase 154: published pages registry
-CREATE TABLE IF NOT EXISTS audit_state.published_pages (
-    id              BIGSERIAL PRIMARY KEY,
-    project_key     TEXT NOT NULL,
-    repo_key        TEXT NOT NULL DEFAULT '',
-    hierarchy_role  TEXT NOT NULL CHECK (hierarchy_role IN (
-        'audit_root', 'analysis_root', 'repo_root', 'repo_run', 'aggregate'
-    )),
-    confluence_page_id   TEXT NOT NULL,
-    confluence_page_title TEXT,
-    parent_page_id       TEXT,
-    space_key            TEXT,
-    created_at      TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    updated_at      TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    CONSTRAINT uq_published_pages_logical_key
-        UNIQUE (project_key, repo_key, hierarchy_role)
-);
-
--- state_repair_hardening: checklist page cache
-CREATE TABLE IF NOT EXISTS audit_state.checklist_cache (
-    id              BIGSERIAL PRIMARY KEY,
-    checklist_page_id TEXT NOT NULL UNIQUE,
-    content_hash    TEXT NOT NULL,
-    parsed_rows     JSONB NOT NULL,
-    row_count       INTEGER NOT NULL,
-    fetched_at      TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    expires_at      TIMESTAMPTZ NOT NULL DEFAULT (NOW() + INTERVAL '24 hours')
-);
-
--- Phase 157: scope snapshots (P157-B durable scope identity)
-CREATE TABLE IF NOT EXISTS audit_state.scope_snapshots (
-    id                    BIGSERIAL PRIMARY KEY,
-    scope_id              TEXT NOT NULL UNIQUE,
-    project_key           TEXT NOT NULL,
-    repo_keys             JSONB NOT NULL,
-    checklist_page_id     TEXT,
-    checklist_row_count   INTEGER,
-    bitbucket_project_keys JSONB,
-    partition_filters     JSONB,
-    metadata_json         JSONB NOT NULL DEFAULT '{}'::jsonb,
-    created_at            TIMESTAMPTZ NOT NULL DEFAULT NOW()
-);
-
--- Phase 157: dispatch state tables (P157-A)
-CREATE TABLE IF NOT EXISTS audit_state.dispatch_runs (
-    id              BIGSERIAL PRIMARY KEY,
-    run_id          TEXT NOT NULL UNIQUE,
-    project_key     TEXT NOT NULL,
-    scope_id        TEXT,
-    status          TEXT NOT NULL DEFAULT 'pending'
-                    CHECK (status IN ('pending', 'running', 'completed', 'failed', 'cancelled')),
-    total_lanes     INTEGER,
-    max_concurrent  INTEGER,
-    sharding_mode   TEXT,
-    operator_overrides_json JSONB NOT NULL DEFAULT '{}'::jsonb,
-    metadata_json   JSONB NOT NULL DEFAULT '{}'::jsonb,
-    started_at      TIMESTAMPTZ,
-    finished_at     TIMESTAMPTZ,
-    heartbeat_at    TIMESTAMPTZ,
-    created_at      TIMESTAMPTZ NOT NULL DEFAULT NOW()
-);
-
-CREATE TABLE IF NOT EXISTS audit_state.dispatch_lanes (
-    id                  BIGSERIAL PRIMARY KEY,
-    dispatch_run_id     BIGINT NOT NULL REFERENCES audit_state.dispatch_runs(id),
-    lane_id             TEXT NOT NULL,
-    profile             TEXT NOT NULL,
-    status              TEXT NOT NULL DEFAULT 'pending'
-                        CHECK (status IN ('pending', 'running', 'completed', 'failed')),
-    repos_planned       INTEGER,
-    repos_completed     INTEGER NOT NULL DEFAULT 0,
-    repos_failed        INTEGER NOT NULL DEFAULT 0,
-    created_at          TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    CONSTRAINT uq_dispatch_lanes_run_profile
-        UNIQUE (dispatch_run_id, profile)
-);
-
--- Phase 168 (P168-A, TSK-168.2): Add claimed_at for stale-claim detection.
--- This was an ALTER in phase157_dispatch_tables.sql; folded in here so a
--- fresh `alembic upgrade head` produces the full Phase 170 schema.
--- Existing DBs: idempotent ADD COLUMN IF NOT EXISTS, harmless on replay.
-ALTER TABLE audit_state.dispatch_lanes
-    ADD COLUMN IF NOT EXISTS claimed_at TIMESTAMPTZ;
-
-CREATE TABLE IF NOT EXISTS audit_state.dispatch_lane_attempts (
-    id                  BIGSERIAL PRIMARY KEY,
-    dispatch_lane_id    BIGINT NOT NULL REFERENCES audit_state.dispatch_lanes(id),
-    repo_key            TEXT NOT NULL,
-    attempt_number      INTEGER NOT NULL DEFAULT 1,
-    status              TEXT NOT NULL DEFAULT 'pending'
-                        CHECK (status IN ('pending', 'running', 'completed', 'failed', 'timeout', 'skipped')),
-    elapsed_seconds     DOUBLE PRECISION,
-    exit_code           INTEGER,
-    error_message       TEXT,
-    metadata_json       JSONB NOT NULL DEFAULT '{}'::jsonb,
-    started_at          TIMESTAMPTZ,
-    finished_at         TIMESTAMPTZ
-);
-
-CREATE TABLE IF NOT EXISTS audit_state.availability_snapshots (
-    id                  BIGSERIAL PRIMARY KEY,
-    dispatch_run_id     BIGINT NOT NULL REFERENCES audit_state.dispatch_runs(id),
-    snapshot_at         TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    available_profiles  JSONB NOT NULL,
-    unavailable_profiles JSONB NOT NULL,
-    metadata_json       JSONB NOT NULL DEFAULT '{}'::jsonb
-);
-
--- Phase 157: Lane contracts (P157-C)
-CREATE TABLE IF NOT EXISTS audit_state.dispatch_lane_contracts (
-    id                  BIGSERIAL PRIMARY KEY,
-    dispatch_lane_id    BIGINT NOT NULL REFERENCES audit_state.dispatch_lanes(id),
-    scope_id            TEXT,
-    profile             TEXT NOT NULL,
-    checklist_page_id   TEXT,
-    target_rows         JSONB,
-    tuning_json         JSONB NOT NULL DEFAULT '{}'::jsonb,
-    contract_json       JSONB NOT NULL DEFAULT '{}'::jsonb,
-    created_at          TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    CONSTRAINT uq_dispatch_lane_contracts_lane
-        UNIQUE (dispatch_lane_id)
-);
-
-CREATE TABLE IF NOT EXISTS audit_state.dispatch_lane_output_envelopes (
-    id                      BIGSERIAL PRIMARY KEY,
-    dispatch_lane_id        BIGINT NOT NULL REFERENCES audit_state.dispatch_lanes(id),
-    attempt_id              BIGINT REFERENCES audit_state.dispatch_lane_attempts(id),
-    status                  TEXT NOT NULL,
-    elapsed_seconds         DOUBLE PRECISION,
-    rows_evaluated          INTEGER,
-    rows_passed             INTEGER,
-    rows_failed             INTEGER,
-    rows_error              INTEGER,
-    evidence_refs_count     INTEGER,
-    envelope_json           JSONB NOT NULL DEFAULT '{}'::jsonb,
-    created_at              TIMESTAMPTZ NOT NULL DEFAULT NOW()
-);
-
--- Phase 157: Dispatch events (P157-D)
-CREATE SEQUENCE IF NOT EXISTS audit_state.dispatch_event_seq;
-
-CREATE TABLE IF NOT EXISTS audit_state.dispatch_events (
-    id                  BIGSERIAL PRIMARY KEY,
-    dispatch_run_id     BIGINT NOT NULL REFERENCES audit_state.dispatch_runs(id),
-    dispatch_lane_id    BIGINT REFERENCES audit_state.dispatch_lanes(id),
-    event_type          TEXT NOT NULL
-                        CHECK (event_type IN (
-                            'run_started', 'run_completed', 'run_failed',
-                            'lane_started', 'lane_completed',
-                            'attempt_started', 'attempt_completed',
-                            'lane_closed_by_doctor', 'dispatch_doctor_repair'
-                        )),
-    event_seq           BIGINT NOT NULL DEFAULT nextval('audit_state.dispatch_event_seq'),
-    payload_json        JSONB NOT NULL DEFAULT '{}'::jsonb,
-    created_at          TIMESTAMPTZ NOT NULL DEFAULT NOW()
-);
-"""
-
-# ─── Phase B: Indexes ────────────────────────────────────────────────────────
-# Mirrors _BOOTSTRAP_INDEXES_SQL in state/store.py.
-# CREATE INDEX (without CONCURRENTLY) is transactional-safe.
-# Operators who need CONCURRENTLY for production rollouts should run those
-# manually; the alembic-managed path uses regular CREATE INDEX (idempotent
-# via IF NOT EXISTS).
-
-_BOOTSTRAP_INDEXES_SQL = """
-CREATE INDEX IF NOT EXISTS idx_audit_state_departments_department_key
-    ON audit_state.departments(department_key);
-CREATE INDEX IF NOT EXISTS idx_audit_state_departments_is_quarantined
-    ON audit_state.departments(is_quarantined);
-CREATE INDEX IF NOT EXISTS idx_audit_state_bitbucket_projects_project_key
-    ON audit_state.bitbucket_projects(project_key);
-CREATE INDEX IF NOT EXISTS idx_audit_state_bitbucket_projects_is_quarantined
-    ON audit_state.bitbucket_projects(is_quarantined);
-CREATE INDEX IF NOT EXISTS idx_audit_state_projects_department_id
-    ON audit_state.projects(department_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_repos_bitbucket_project_id
-    ON audit_state.repos(bitbucket_project_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_repos_project_id
-    ON audit_state.repos(project_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_repos_project_analysis_excluded
-    ON audit_state.repos(project_id, analysis_excluded);
-CREATE INDEX IF NOT EXISTS idx_audit_state_artifacts_repo_id
-    ON audit_state.artifacts(repo_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_artifacts_status
-    ON audit_state.artifacts(status);
-CREATE INDEX IF NOT EXISTS idx_audit_state_artifacts_claim_expiry
-    ON audit_state.artifacts(claim_expires_at);
-CREATE INDEX IF NOT EXISTS idx_audit_state_runs_project_key
-    ON audit_state.runs(project_key);
-CREATE INDEX IF NOT EXISTS idx_audit_state_runs_repo_key
-    ON audit_state.runs(repo_key);
-CREATE INDEX IF NOT EXISTS idx_audit_state_runs_status
-    ON audit_state.runs(status);
-CREATE INDEX IF NOT EXISTS idx_audit_state_profile_performance_profile_project_created
-    ON audit_state.profile_performance(profile_name, project_key, created_at DESC);
-CREATE INDEX IF NOT EXISTS idx_audit_state_run_events_run_pk
-    ON audit_state.run_events(run_pk);
-CREATE INDEX IF NOT EXISTS idx_audit_state_checklist_runs_project_key
-    ON audit_state.checklist_runs(project_key);
-CREATE INDEX IF NOT EXISTS idx_audit_state_checklist_runs_repo_key
-    ON audit_state.checklist_runs(repo_key);
-CREATE INDEX IF NOT EXISTS idx_audit_state_checklist_runs_thread_id
-    ON audit_state.checklist_runs(thread_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_checklist_runs_status
-    ON audit_state.checklist_runs(status);
-CREATE INDEX IF NOT EXISTS idx_audit_state_checklist_row_results_checklist_run_id
-    ON audit_state.checklist_row_results(checklist_run_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_checklist_row_results_row_index
-    ON audit_state.checklist_row_results(row_index);
-CREATE INDEX IF NOT EXISTS idx_audit_state_checklist_row_results_status
-    ON audit_state.checklist_row_results(status);
-CREATE INDEX IF NOT EXISTS idx_audit_state_checklist_artifacts_checklist_run_id
-    ON audit_state.checklist_artifacts(checklist_run_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_checklist_artifacts_artifact_hash
-    ON audit_state.checklist_artifacts(artifact_hash);
-CREATE INDEX IF NOT EXISTS idx_audit_state_evidence_snapshots_project_key
-    ON audit_state.evidence_snapshots(project_key);
-CREATE INDEX IF NOT EXISTS idx_audit_state_evidence_snapshots_repo_key
-    ON audit_state.evidence_snapshots(repo_key);
-CREATE INDEX IF NOT EXISTS idx_audit_state_evidence_snapshots_content_hash
-    ON audit_state.evidence_snapshots(content_hash);
-CREATE INDEX IF NOT EXISTS idx_audit_state_evidence_snapshots_project_created
-    ON audit_state.evidence_snapshots(project_key, created_at DESC);
-CREATE INDEX IF NOT EXISTS idx_audit_state_retrieval_snapshots_project_key
-    ON audit_state.retrieval_snapshots(project_key);
-CREATE INDEX IF NOT EXISTS idx_audit_state_retrieval_snapshots_repo_key
-    ON audit_state.retrieval_snapshots(repo_key);
-CREATE INDEX IF NOT EXISTS idx_audit_state_retrieval_snapshots_project_created
-    ON audit_state.retrieval_snapshots(project_key, created_at DESC);
-CREATE INDEX IF NOT EXISTS idx_audit_state_project_links_project_id
-    ON audit_state.project_links(project_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_project_links_link_type_role
-    ON audit_state.project_links(link_type, link_role);
-CREATE INDEX IF NOT EXISTS idx_audit_state_project_links_normalized_url
-    ON audit_state.project_links(normalized_url);
-CREATE INDEX IF NOT EXISTS idx_audit_state_project_links_source_page_id
-    ON audit_state.project_links(source_page_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_project_links_last_seen_at
-    ON audit_state.project_links(last_seen_at DESC);
-CREATE INDEX IF NOT EXISTS idx_audit_state_repo_links_repo_id
-    ON audit_state.repo_links(repo_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_repo_links_link_type_role
-    ON audit_state.repo_links(link_type, link_role);
-CREATE INDEX IF NOT EXISTS idx_audit_state_repo_links_normalized_url
-    ON audit_state.repo_links(normalized_url);
-CREATE INDEX IF NOT EXISTS idx_audit_state_repo_links_source_page_id
-    ON audit_state.repo_links(source_page_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_repo_links_last_seen_at
-    ON audit_state.repo_links(last_seen_at DESC);
-CREATE INDEX IF NOT EXISTS idx_audit_state_project_preparation_project_id
-    ON audit_state.project_preparation(project_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_repo_preparation_repo_id
-    ON audit_state.repo_preparation(repo_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_corpus_documents_snapshot_id
-    ON audit_state.corpus_documents(snapshot_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_corpus_documents_project_id
-    ON audit_state.corpus_documents(project_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_corpus_documents_repo_id
-    ON audit_state.corpus_documents(repo_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_corpus_documents_parent_document_id
-    ON audit_state.corpus_documents(parent_document_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_corpus_documents_source_type_doc_class
-    ON audit_state.corpus_documents(source_type, doc_class);
-CREATE INDEX IF NOT EXISTS idx_audit_state_corpus_documents_content_hash
-    ON audit_state.corpus_documents(content_hash);
-CREATE INDEX IF NOT EXISTS idx_audit_state_corpus_documents_project_source_uri
-    ON audit_state.corpus_documents(project_id, source_uri);
-CREATE INDEX IF NOT EXISTS idx_audit_state_corpus_documents_project_space_key
-    ON audit_state.corpus_documents(project_id, space_key);
-CREATE INDEX IF NOT EXISTS idx_audit_state_corpus_chunks_document_id
-    ON audit_state.corpus_chunks(document_pk);
-CREATE INDEX IF NOT EXISTS idx_audit_state_corpus_chunks_search_vector
-    ON audit_state.corpus_chunks
-    USING GIN(search_vector);
-CREATE INDEX IF NOT EXISTS idx_audit_state_corpus_chunks_content_hash
-    ON audit_state.corpus_chunks(content_hash);
-CREATE INDEX IF NOT EXISTS idx_audit_state_corpus_embeddings_snapshot_id
-    ON audit_state.corpus_embeddings(snapshot_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_corpus_embeddings_chunk_pk
-    ON audit_state.corpus_embeddings(chunk_pk);
-CREATE INDEX IF NOT EXISTS idx_audit_state_project_bitbucket_links_project_id
-    ON audit_state.project_bitbucket_links(project_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_project_bitbucket_links_bitbucket_project_id
-    ON audit_state.project_bitbucket_links(bitbucket_project_id);
-
--- Phase 114: requirement_classifications indexes
-CREATE INDEX IF NOT EXISTS idx_audit_state_requirement_classifications_row_key
-    ON audit_state.requirement_classifications(row_key);
-CREATE INDEX IF NOT EXISTS idx_audit_state_requirement_classifications_content_fingerprint
-    ON audit_state.requirement_classifications(content_fingerprint);
-CREATE INDEX IF NOT EXISTS idx_audit_state_requirement_classifications_project_run
-    ON audit_state.requirement_classifications(project_storage_key, run_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_requirement_classifications_checklist_page
-    ON audit_state.requirement_classifications(checklist_page_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_requirement_classifications_classification_source
-    ON audit_state.requirement_classifications(classification_source);
-
--- Phase 117/118: shared state indexes
-CREATE INDEX IF NOT EXISTS idx_audit_state_project_evidence_discoveries_context_id
-    ON audit_state.project_evidence_discoveries(context_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_project_evidence_discoveries_source_repo
-    ON audit_state.project_evidence_discoveries(source_repo);
-CREATE INDEX IF NOT EXISTS idx_audit_state_project_evidence_discoveries_discovery_type
-    ON audit_state.project_evidence_discoveries(discovery_type);
-CREATE INDEX IF NOT EXISTS idx_audit_state_project_evidence_discoveries_context_created_at
-    ON audit_state.project_evidence_discoveries(context_id, created_at DESC);
-CREATE INDEX IF NOT EXISTS idx_audit_state_project_knowledge_project_repo_type_created
-    ON audit_state.project_knowledge(project_key, repo_key, knowledge_type, created_at DESC);
-CREATE INDEX IF NOT EXISTS idx_audit_state_project_knowledge_project_created_at
-    ON audit_state.project_knowledge(project_key, created_at DESC);
-CREATE INDEX IF NOT EXISTS idx_audit_state_project_knowledge_expires_at
-    ON audit_state.project_knowledge(expires_at);
-CREATE INDEX IF NOT EXISTS idx_audit_state_project_knowledge_source_run_id
-    ON audit_state.project_knowledge(source_run_id);
-
--- Phase 154: published pages indexes
-CREATE INDEX IF NOT EXISTS ix_published_pages_project_key
-    ON audit_state.published_pages(project_key);
-CREATE INDEX IF NOT EXISTS ix_published_pages_confluence_page_id
-    ON audit_state.published_pages(confluence_page_id);
-
--- state_repair_hardening
-CREATE INDEX IF NOT EXISTS ix_checklist_cache_page_id
-    ON audit_state.checklist_cache(checklist_page_id);
-
--- Phase 157: scope_snapshots indexes (P157-B)
-CREATE INDEX IF NOT EXISTS ix_scope_snapshots_project_key
-    ON audit_state.scope_snapshots(project_key);
-CREATE INDEX IF NOT EXISTS ix_scope_snapshots_created_at
-    ON audit_state.scope_snapshots(created_at DESC);
-
--- Phase 157: dispatch state indexes (P157-A)
-CREATE INDEX IF NOT EXISTS idx_audit_state_dispatch_runs_project_key
-    ON audit_state.dispatch_runs(project_key);
-CREATE INDEX IF NOT EXISTS idx_audit_state_dispatch_runs_status
-    ON audit_state.dispatch_runs(status);
-CREATE INDEX IF NOT EXISTS idx_audit_state_dispatch_lanes_dispatch_run_id
-    ON audit_state.dispatch_lanes(dispatch_run_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_dispatch_lanes_status
-    ON audit_state.dispatch_lanes(status);
-CREATE INDEX IF NOT EXISTS idx_audit_state_dispatch_lane_attempts_dispatch_lane_id
-    ON audit_state.dispatch_lane_attempts(dispatch_lane_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_dispatch_lane_attempts_status
-    ON audit_state.dispatch_lane_attempts(status);
-CREATE INDEX IF NOT EXISTS idx_audit_state_availability_snapshots_dispatch_run_id
-    ON audit_state.availability_snapshots(dispatch_run_id);
-
--- Phase 157: lane contract indexes (P157-C)
-CREATE INDEX IF NOT EXISTS idx_audit_state_dispatch_lane_contracts_lane_id
-    ON audit_state.dispatch_lane_contracts(dispatch_lane_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_dispatch_lane_contracts_profile
-    ON audit_state.dispatch_lane_contracts(profile);
-CREATE INDEX IF NOT EXISTS idx_audit_state_dispatch_lane_output_envelopes_lane_id
-    ON audit_state.dispatch_lane_output_envelopes(dispatch_lane_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_dispatch_lane_output_envelopes_attempt_id
-    ON audit_state.dispatch_lane_output_envelopes(attempt_id);
-
--- Phase 157: dispatch event indexes (P157-D)
-CREATE INDEX IF NOT EXISTS idx_audit_state_dispatch_events_run_id
-    ON audit_state.dispatch_events(dispatch_run_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_dispatch_events_lane_id
-    ON audit_state.dispatch_events(dispatch_lane_id);
-CREATE INDEX IF NOT EXISTS idx_audit_state_dispatch_events_event_type
-    ON audit_state.dispatch_events(event_type);
-CREATE INDEX IF NOT EXISTS idx_audit_state_dispatch_events_event_seq
-    ON audit_state.dispatch_events(event_seq);
-"""
-
-# P4 v2.12.1 fix: ensure the thread_id column exists on audit_state.runs before
-# creating the partial index. The legacy `_reconcile_bootstrap_columns` runtime
-# helper in state/store.py added this column at runtime; the squash must
-# materialize it so a fresh-deploy alembic upgrade head succeeds.
-# Partial index — idempotent.
-_BOOTSTRAP_PARTIAL_INDEX_SQL = """
-ALTER TABLE audit_state.runs
-    ADD COLUMN IF NOT EXISTS thread_id TEXT;
-
-CREATE INDEX IF NOT EXISTS idx_audit_state_runs_thread_id
-    ON audit_state.runs (thread_id) WHERE thread_id IS NOT NULL;
-"""
-
-
-# ─── Phase C: SQL functions + view ───────────────────────────────────────────
-# Mirrors _BOOTSTRAP_FUNCTIONS_SQL in state/store.py.
-
-_BOOTSTRAP_FUNCTIONS_SQL = """
--- Phase 47: RRF Hybrid Retrieval Functions (FR-47.2.5)
-CREATE OR REPLACE FUNCTION audit_state.rrf_score(rank INTEGER, rrf_k INTEGER DEFAULT 60)
-RETURNS NUMERIC
-LANGUAGE SQL
-IMMUTABLE
-PARALLEL SAFE
-AS $$
-    SELECT CASE
-        WHEN rank IS NULL OR rank < 1 THEN 0
-        ELSE 1.0 / (rrf_k + rank)
-    END
-$$;
-
--- Hybrid search combining lexical (tsvector) and vector similarity with RRF fusion.
-CREATE OR REPLACE FUNCTION audit_state.hybrid_search(
-    p_project_key TEXT,
-    p_repo_key TEXT DEFAULT NULL,
-    p_query_text TEXT DEFAULT '',
-    p_query_embedding vector DEFAULT NULL,
-    p_top_k INTEGER DEFAULT 10,
-    p_lexical_weight NUMERIC DEFAULT 0.4,
-    p_vector_weight NUMERIC DEFAULT 0.6,
-    p_snapshot_id BIGINT DEFAULT NULL
-)
-RETURNS TABLE (
-    chunk_pk BIGINT,
-    document_id TEXT,
-    chunk_id TEXT,
-    content_text TEXT,
-    source_uri TEXT,
-    page_id TEXT,
-    content_hash TEXT,
-    metadata JSONB,
-    lexical_rank BIGINT,
-    vector_rank BIGINT,
-    rrf_score NUMERIC,
-    page_title TEXT,
-    space_key TEXT,
-    source_type TEXT,
-    doc_class TEXT
-)
-LANGUAGE plpgsql
-STABLE
-AS $$
-DECLARE
-    v_candidate_limit INTEGER := GREATEST(p_top_k * 3, 30);
-    v_snapshot_id BIGINT := p_snapshot_id;
-    v_project_id BIGINT;
-    v_repo_id BIGINT;
-BEGIN
-    SELECT id INTO v_project_id
-    FROM audit_state.projects
-    WHERE project_key = p_project_key
-    LIMIT 1;
-
-    IF v_project_id IS NULL THEN
-        RETURN;
-    END IF;
-
-    IF p_repo_key IS NOT NULL THEN
-        SELECT id INTO v_repo_id
-        FROM audit_state.repos
-        WHERE repo_key = p_repo_key
-          AND project_id = v_project_id
-        LIMIT 1;
-    END IF;
-
-    IF v_snapshot_id IS NULL THEN
-        SELECT rs.id INTO v_snapshot_id
-        FROM audit_state.retrieval_snapshots rs
-        WHERE rs.project_key = p_project_key
-          AND (
-              (p_repo_key IS NULL AND rs.repo_key IS NULL)
-              OR (p_repo_key IS NOT NULL AND rs.repo_key = p_repo_key)
-          )
-        ORDER BY
-            CASE
-                WHEN lower(rs.snapshot_key) LIKE '%\\:project_corpus' THEN 3
-                WHEN lower(rs.snapshot_key) LIKE '%\\:smoke_probe' THEN 1
-                ELSE 2
-            END DESC,
-            COALESCE(rs.chunk_count, 0) DESC,
-            COALESCE(rs.changed_chunk_count, 0) DESC,
-            rs.created_at DESC,
-            rs.id DESC
-        LIMIT 1;
-    END IF;
-
-    RETURN QUERY
-    WITH scoped_chunks AS (
-        SELECT
-            cc.id AS chunk_pk,
-            cc.chunk_id,
-            cc.content_text,
-            cc.page_id,
-            cc.content_hash,
-            cc.metadata,
-            cd.document_id,
-            cd.source_uri,
-            cd.page_title,
-            cd.space_key,
-            cd.source_type,
-            cd.doc_class
-        FROM audit_state.corpus_chunks cc
-        JOIN audit_state.corpus_documents cd ON cd.id = cc.document_pk
-        WHERE cd.project_id = v_project_id
-          AND (v_repo_id IS NULL OR cd.repo_id = v_repo_id OR (cd.repo_id IS NULL AND cd.source_type NOT IN ('code_file', 'ops_repo_file')))
-          AND (v_snapshot_id IS NULL OR cd.snapshot_id = v_snapshot_id)
-    ),
-    lexical_results AS (
-        SELECT
-            sc.chunk_pk,
-            ROW_NUMBER() OVER (
-                ORDER BY ts_rank_cd(sc.search_vector, plainto_tsquery('simple', p_query_text)) DESC
-            ) AS rank
-        FROM (
-            SELECT cc.id AS chunk_pk, cc.chunk_id, cc.content_text,
-                   to_tsvector('simple', COALESCE(cc.content_text, '')) AS search_vector
-            FROM audit_state.corpus_chunks cc
-            JOIN audit_state.corpus_documents cd ON cd.id = cc.document_pk
-            WHERE cd.project_id = v_project_id
-              AND (v_repo_id IS NULL OR cd.repo_id = v_repo_id OR (cd.repo_id IS NULL AND cd.source_type NOT IN ('code_file', 'ops_repo_file')))
-              AND (v_snapshot_id IS NULL OR cd.snapshot_id = v_snapshot_id)
-        ) sc
-        WHERE sc.search_vector @@ plainto_tsquery('simple', p_query_text)
-        ORDER BY ts_rank_cd(sc.search_vector, plainto_tsquery('simple', p_query_text)) DESC
-        LIMIT v_candidate_limit
-    ),
-    vector_results AS (
-        SELECT
-            ce.chunk_pk,
-            ROW_NUMBER() OVER (ORDER BY ce.embedding_vector <=> p_query_embedding) AS rank
-        FROM audit_state.corpus_embeddings ce
-        JOIN audit_state.corpus_documents cd ON cd.snapshot_id = ce.snapshot_id
-        WHERE cd.project_id = v_project_id
-          AND (v_repo_id IS NULL OR cd.repo_id = v_repo_id OR (cd.repo_id IS NULL AND cd.source_type NOT IN ('code_file', 'ops_repo_file')))
-          AND (v_snapshot_id IS NULL OR ce.snapshot_id = v_snapshot_id)
-          AND ce.embedding_vector IS NOT NULL
-          AND p_query_embedding IS NOT NULL
-        ORDER BY ce.embedding_vector <=> p_query_embedding
-        LIMIT v_candidate_limit
-    ),
-    fused AS (
-        SELECT
-            COALESCE(l.chunk_pk, v.chunk_pk) AS chunk_pk,
-            COALESCE(audit_state.rrf_score(l.rank::INTEGER), 0) * p_lexical_weight
-            + COALESCE(audit_state.rrf_score(v.rank::INTEGER), 0) * p_vector_weight AS combined_score,
-            l.rank AS lexical_rank,
-            v.rank AS vector_rank
-        FROM lexical_results l
-        FULL OUTER JOIN vector_results v ON l.chunk_pk = v.chunk_pk
-    )
-    SELECT
-        f.chunk_pk,
-        sc.document_id,
-        sc.chunk_id,
-        sc.content_text,
-        sc.source_uri,
-        sc.page_id,
-        sc.content_hash,
-        sc.metadata,
-        f.lexical_rank,
-        f.vector_rank,
-        f.combined_score AS rrf_score,
-        sc.page_title,
-        sc.space_key,
-        sc.source_type,
-        sc.doc_class
-    FROM fused f
-    JOIN scoped_chunks sc ON sc.chunk_pk = f.chunk_pk
-    ORDER BY f.combined_score DESC
-    LIMIT p_top_k;
-END;
-$$;
-
--- Phase 47: Advisory Lock Functions (FR-47.2.10)
-CREATE OR REPLACE FUNCTION audit_state.acquire_corpus_lock(
-    p_project_storage_key TEXT,
-    p_timeout_ms INTEGER DEFAULT 5000
-)
-RETURNS BOOLEAN
-LANGUAGE plpgsql
-AS $$
-DECLARE
-    v_lock_key BIGINT;
-    v_acquired BOOLEAN := FALSE;
-    v_start_ts TIMESTAMPTZ := clock_timestamp();
-    v_elapsed_ms NUMERIC;
-BEGIN
-    v_lock_key := hashtext('corpus:' || p_project_storage_key);
-
-    LOOP
-        SELECT pg_try_advisory_xact_lock(v_lock_key) INTO v_acquired;
-
-        IF v_acquired THEN
-            RETURN TRUE;
-        END IF;
-
-        v_elapsed_ms := EXTRACT(MILLISECONDS FROM clock_timestamp() - v_start_ts);
-        IF v_elapsed_ms >= p_timeout_ms THEN
-            RAISE WARNING 'Corpus lock acquisition timeout for project=% after %ms',
-                p_project_storage_key, p_timeout_ms;
-            RETURN FALSE;
-        END IF;
-
-        PERFORM pg_sleep(0.1);
-    END LOOP;
-END;
-$$;
-
-CREATE OR REPLACE FUNCTION audit_state.acquire_repo_sync_lock(
-    p_repo_storage_key TEXT,
-    p_timeout_ms INTEGER DEFAULT 5000
-)
-RETURNS BOOLEAN
-LANGUAGE plpgsql
-AS $$
-DECLARE
-    v_lock_key BIGINT;
-    v_acquired BOOLEAN := FALSE;
-    v_start_ts TIMESTAMPTZ := clock_timestamp();
-    v_elapsed_ms NUMERIC;
-BEGIN
-    v_lock_key := hashtext('repo_sync:' || p_repo_storage_key);
-
-    LOOP
-        SELECT pg_try_advisory_xact_lock(v_lock_key) INTO v_acquired;
-
-        IF v_acquired THEN
-            RETURN TRUE;
-        END IF;
-
-        v_elapsed_ms := EXTRACT(MILLISECONDS FROM clock_timestamp() - v_start_ts);
-        IF v_elapsed_ms >= p_timeout_ms THEN
-            RAISE WARNING 'Repo sync lock acquisition timeout for repo=% after %ms',
-                p_repo_storage_key, p_timeout_ms;
-            RETURN FALSE;
-        END IF;
-
-        PERFORM pg_sleep(0.1);
-    END LOOP;
-END;
-$$;
-
--- Phase 47: Vietnamese Accent-Insensitive Search (FR-47.9.7)
-CREATE OR REPLACE FUNCTION audit_state.vds_search(
-    p_query TEXT,
-    p_limit INTEGER DEFAULT 10
-) RETURNS TABLE (
-    chunk_pk BIGINT,
-    document_pk BIGINT,
-    content_text TEXT,
-    rank REAL
-) LANGUAGE plpgsql STABLE AS $$
-BEGIN
-    RETURN QUERY
-    SELECT
-        cc.id AS chunk_pk,
-        cc.document_pk,
-        cc.content_text,
-        ts_rank_cd(cc.search_vector, to_tsquery('simple', unaccent(p_query))) AS rank
-    FROM audit_state.corpus_chunks cc
-    WHERE cc.search_vector @@ to_tsquery('simple', unaccent(p_query))
-    ORDER BY rank DESC
-    LIMIT p_limit;
-END;
-$$;
-
--- Phase 157/169: Dispatch Run Status View
--- DROP first because CREATE OR REPLACE VIEW cannot change column names on
--- an existing view (Phase 169 added heartbeat_at to an older Phase 157 view).
-DROP VIEW IF EXISTS audit_state.dispatch_run_status;
-CREATE VIEW audit_state.dispatch_run_status AS
-SELECT
-    dr.run_id,
-    dr.project_key,
-    dr.status AS run_status,
-    dr.started_at,
-    dr.finished_at,
-    dr.heartbeat_at,
-    COUNT(dl.id) AS total_lanes,
-    COUNT(CASE WHEN dl.status = 'completed' THEN 1 END) AS completed_lanes,
-    COUNT(CASE WHEN dl.status = 'failed' THEN 1 END) AS failed_lanes,
-    COUNT(CASE WHEN dl.status = 'running' THEN 1 END) AS active_lanes,
-    COALESCE(SUM(dl.repos_completed), 0) AS total_repos_completed,
-    COALESCE(SUM(dl.repos_failed), 0) AS total_repos_failed
-FROM audit_state.dispatch_runs dr
-LEFT JOIN audit_state.dispatch_lanes dl ON dl.dispatch_run_id = dr.id
-GROUP BY dr.id;
-"""
-
-
-# ─── schema_version seed (Phase 47/50, version=52) ───────────────────────────
-# Mirrors the INSERT inside init_schema(). Idempotent via ON CONFLICT.
-_SCHEMA_VERSION_SEED_SQL = """
-INSERT INTO audit_state.schema_version (id, version, phase, applied_at)
-VALUES (1, 52, 'Phase 50', NOW())
-ON CONFLICT (id) DO UPDATE
-SET
-    version = EXCLUDED.version,
-    phase = EXCLUDED.phase,
-    applied_at = EXCLUDED.applied_at;
-"""
-
-
-def upgrade() -> None:
-    # Extensions (vector, unaccent) — must come before any vector/unaccent usage.
-    op.execute(sa.text(_BOOTSTRAP_EXTENSIONS_SQL))
-
-    # Tables + constraints + the Phase 168 dispatch_lanes.claimed_at ALTER.
-    op.execute(sa.text(_BOOTSTRAP_TABLES_SQL))
-
-    # Indexes (regular CREATE INDEX, transactional-safe; idempotent).
-    op.execute(sa.text(_BOOTSTRAP_INDEXES_SQL))
-    op.execute(sa.text(_BOOTSTRAP_PARTIAL_INDEX_SQL))
-
-    # Functions, advisory locks, and dispatch_run_status view.
-    op.execute(sa.text(_BOOTSTRAP_FUNCTIONS_SQL))
-
-    # Seed schema_version anchor (idempotent).
-    op.execute(sa.text(_SCHEMA_VERSION_SEED_SQL))
-
-
-def downgrade() -> None:
-    # Schema retirement is a manual operator decision, NOT an alembic concern.
-    # We do NOT drop audit_state CASCADE — that would destroy years of audit
-    # state, evidence snapshots, retrieval embeddings, and dispatch history.
-    # Operators retire schemas via dedicated runbooks, not via alembic.
-    pass