politics-tracking/backend/migrations/0001_schema.sql

-- Politrack schema (v2). See docs/architecture/01-design-from-scratch.md and docs/data-model.md.
--
-- Every user-visible fact is a CLAIM backed by >= 1 EVIDENCE row pointing to a DOCUMENT (a URL)
-- published by a SOURCE. Typed facet tables (role, vote_position, money_flow, ...) add structure.
-- Entities carry any number of external IDENTIFIERS. Raw responses are indexed in FETCH.
-- All objects carry COMMENTs so agents can discover semantics from the catalog.

CREATE EXTENSION IF NOT EXISTS vector;
CREATE EXTENSION IF NOT EXISTS pg_trgm;
CREATE EXTENSION IF NOT EXISTS unaccent;

-- ---------------------------------------------------------------------------
-- Sources, raw archive, documents
-- ---------------------------------------------------------------------------

CREATE TABLE source (
    id              SERIAL PRIMARY KEY,
    slug            TEXT NOT NULL UNIQUE,
    name            TEXT NOT NULL,
    kind            TEXT NOT NULL CHECK (kind IN ('official', 'ngo', 'media', 'social', 'aggregator', 'user', 'derived')),
    homepage        TEXT,
    license         TEXT,
    access_method   TEXT NOT NULL CHECK (access_method IN ('api', 'bulk', 'rss', 'scrape', 'manual', 'derived')),
    store_fulltext  BOOLEAN NOT NULL DEFAULT FALSE,
    keep_raw        BOOLEAN NOT NULL DEFAULT TRUE,
    default_trust   SMALLINT NOT NULL DEFAULT 50 CHECK (default_trust BETWEEN 0 AND 100),
    tos_notes       TEXT,
    updated_at      TIMESTAMPTZ NOT NULL DEFAULT now()
);
COMMENT ON TABLE source IS 'A publisher (parliament API, a newspaper, a social network...). Rows are synced from the SourceSpec objects in backend/politrack/sources (single source of truth).';
COMMENT ON COLUMN source.store_fulltext IS 'FALSE when licence/ToS or a TDM reservation forbids keeping body text: only title, short excerpt and link are stored.';
COMMENT ON COLUMN source.keep_raw IS 'FALSE when raw responses must not be archived (e.g. TDM-reserved publishers).';

CREATE TABLE fetch_log (
    id              BIGSERIAL PRIMARY KEY,
    adapter         TEXT NOT NULL,
    source_id       INT REFERENCES source(id),
    request_fp      TEXT NOT NULL,
    url             TEXT NOT NULL,
    method          TEXT NOT NULL DEFAULT 'GET',
    request         JSONB NOT NULL DEFAULT '{}'::jsonb,
    status          INT,
    sha256          TEXT,
    size            INT,
    content_type    TEXT,
    etag            TEXT,
    last_modified   TEXT,
    stored          BOOLEAN NOT NULL DEFAULT FALSE,
    tdm_reserved    BOOLEAN NOT NULL DEFAULT FALSE,
    fetched_at      TIMESTAMPTZ NOT NULL DEFAULT now(),
    parsed_sha256   TEXT,
    parser_version  INT,
    error           TEXT
);
CREATE INDEX fetch_fp_idx ON fetch_log (request_fp, fetched_at DESC);
CREATE INDEX fetch_adapter_idx ON fetch_log (adapter, fetched_at DESC);
COMMENT ON TABLE fetch_log IS 'Raw archive index: one row per HTTP fetch. Bodies live gzip-compressed under POLITRACK_RAW_DIR/<sha256[:2]>/<sha256>.gz when stored=TRUE. Used for conditional requests, change detection and re-parsing.';

CREATE TABLE story (
    id              BIGSERIAL PRIMARY KEY,
    title           TEXT NOT NULL,
    first_seen      TIMESTAMPTZ,
    last_seen       TIMESTAMPTZ,
    document_count  INT NOT NULL DEFAULT 0,
    source_count    INT NOT NULL DEFAULT 0
);
COMMENT ON TABLE story IS 'Cluster of news documents about the same event. Reported by several independent outlets = corroborated.';

CREATE TABLE document (
    id              BIGSERIAL PRIMARY KEY,
    source_id       INT NOT NULL REFERENCES source(id),
    url             TEXT NOT NULL UNIQUE,
    title           TEXT,
    doc_type        TEXT NOT NULL DEFAULT 'web_page',
    lang            TEXT DEFAULT 'de',
    published_at    TIMESTAMPTZ,
    retrieved_at    TIMESTAMPTZ NOT NULL DEFAULT now(),
    excerpt         TEXT,
    full_text       TEXT,
    archive_url     TEXT,
    fetch_id        BIGINT REFERENCES fetch_log(id) ON DELETE SET NULL,
    story_id        BIGINT REFERENCES story(id) ON DELETE SET NULL,
    meta            JSONB NOT NULL DEFAULT '{}'::jsonb
);
CREATE INDEX document_source_idx ON document (source_id, published_at DESC);
CREATE INDEX document_published_idx ON document (published_at DESC);
CREATE INDEX document_story_idx ON document (story_id);
CREATE INDEX document_title_trgm ON document USING gin (title gin_trgm_ops);
COMMENT ON TABLE document IS 'One URL (article, API record, protocol, post, register entry). Evidence rows point here; url is what users click to verify a claim.';
COMMENT ON COLUMN document.doc_type IS 'article | press_release | api_record | protocol | social_post | register_entry | web_page | pdf | dataset_row';

-- ---------------------------------------------------------------------------
-- Entities
-- ---------------------------------------------------------------------------

CREATE TABLE person (
    id              SERIAL PRIMARY KEY,
    slug            TEXT NOT NULL UNIQUE,
    full_name       TEXT NOT NULL,
    given_name      TEXT,
    family_name     TEXT,
    honorific_prefix TEXT,
    honorific_suffix TEXT,
    birth_date      DATE,
    death_date      DATE,
    gender          TEXT,
    photo_url       TEXT,
    summary_de      TEXT,
    summary_en      TEXT,
    created_at      TIMESTAMPTZ NOT NULL DEFAULT now(),
    updated_at      TIMESTAMPTZ NOT NULL DEFAULT now()
);
CREATE INDEX person_name_trgm ON person USING gin (full_name gin_trgm_ops);
COMMENT ON TABLE person IS 'A politician: anyone holding or having held a party, parliamentary, government or municipal function.';

CREATE TABLE organization (
    id              SERIAL PRIMARY KEY,
    slug            TEXT NOT NULL UNIQUE,
    name            TEXT NOT NULL,
    short_name      TEXT,
    kind            TEXT NOT NULL CHECK (kind IN ('party', 'parliamentary_group', 'parliament', 'government', 'ministry',
                                                  'state_government', 'municipality', 'public_body', 'company', 'ngo',
                                                  'media', 'lobby', 'court', 'authority', 'eu_institution', 'other')),
    parent_id       INT REFERENCES organization(id),
    color           TEXT,
    logo_url        TEXT,
    created_at      TIMESTAMPTZ NOT NULL DEFAULT now()
);
CREATE INDEX organization_name_trgm ON organization USING gin (name gin_trgm_ops);
COMMENT ON TABLE organization IS 'Parties, parliamentary groups, chambers, ministries, public bodies, companies, media owners, lobby organisations.';

CREATE TABLE identifier (
    entity_type     TEXT NOT NULL CHECK (entity_type IN ('person', 'organization')),
    entity_id       INT NOT NULL,
    scheme          TEXT NOT NULL,
    value           TEXT NOT NULL,
    source_slug     TEXT,
    verified        BOOLEAN NOT NULL DEFAULT TRUE,
    PRIMARY KEY (entity_type, scheme, value)
);
CREATE INDEX identifier_entity_idx ON identifier (entity_type, entity_id);
COMMENT ON TABLE identifier IS 'External ids: scheme parlament (PAD_INTERN), wikidata (QID), ep (MEP id), bluesky (DID), mastodon (acct), telegram (channel), youtube (channel id), uid (ATU...), firmenbuch, rtr (Rechtsträger id), eu_transparency (register id)...';

CREATE TABLE alias (
    entity_type     TEXT NOT NULL CHECK (entity_type IN ('person', 'organization')),
    entity_id       INT NOT NULL,
    alias           TEXT NOT NULL,
    alias_norm      TEXT NOT NULL,
    PRIMARY KEY (entity_type, entity_id, alias_norm)
);
CREATE INDEX alias_norm_idx ON alias (entity_type, alias_norm);
COMMENT ON TABLE alias IS 'Name variants for entity resolution and linking (titles stripped, legal forms normalised for organisations).';

CREATE TABLE social_account (
    id              SERIAL PRIMARY KEY,
    person_id       INT REFERENCES person(id) ON DELETE CASCADE,
    organization_id INT REFERENCES organization(id) ON DELETE CASCADE,
    platform        TEXT NOT NULL,
    handle          TEXT,
    url             TEXT NOT NULL,
    verified        BOOLEAN NOT NULL DEFAULT FALSE,
    source_slug     TEXT,
    UNIQUE (platform, url)
);
COMMENT ON COLUMN social_account.verified IS 'TRUE when the account comes from an official source (parliament profile, party/official website, platform verification); Wikidata-only handles are FALSE until confirmed.';

-- ---------------------------------------------------------------------------
-- Topics, claims, evidence
-- ---------------------------------------------------------------------------

CREATE TABLE topic (
    id              SERIAL PRIMARY KEY,
    slug            TEXT NOT NULL UNIQUE,
    name_de         TEXT NOT NULL,
    name_en         TEXT NOT NULL,
    keywords        TEXT[] NOT NULL DEFAULT '{}'
);
COMMENT ON TABLE topic IS 'Bilingual topic taxonomy; the first 18 mirror the Austrian Parliament''s THEMEN.';

CREATE TABLE claim (
    id              BIGSERIAL PRIMARY KEY,
    claim_type      TEXT NOT NULL CHECK (claim_type IN (
                        'role', 'vote', 'donation', 'money_flow', 'statement', 'appearance', 'meeting', 'travel',
                        'side_job', 'legal_proceeding', 'lobbying', 'media_mention', 'career',
                        'parliamentary_question', 'government_decision', 'wiki_edit', 'declaration', 'ownership',
                        'affiliation', 'party_finance', 'other')),
    fingerprint     TEXT NOT NULL UNIQUE,
    person_id       INT REFERENCES person(id) ON DELETE CASCADE,
    organization_id INT REFERENCES organization(id) ON DELETE CASCADE,
    counterpart_person_id INT REFERENCES person(id),
    counterpart_org_id    INT REFERENCES organization(id),
    occurred_at     TIMESTAMPTZ,
    valid_from      DATE,
    valid_to        DATE,
    summary_de      TEXT NOT NULL,
    summary_en      TEXT,
    data            JSONB NOT NULL DEFAULT '{}'::jsonb,
    status          TEXT NOT NULL DEFAULT 'reported' CHECK (status IN ('reported', 'corroborated', 'disputed', 'retracted')),
    sensitive       BOOLEAN NOT NULL DEFAULT FALSE,
    review_status   TEXT NOT NULL DEFAULT 'published' CHECK (review_status IN ('pending', 'published', 'rejected')),
    reviewed_by     TEXT,
    reviewed_at     TIMESTAMPTZ,
    merged_into     BIGINT REFERENCES claim(id) ON DELETE SET NULL,
    created_by      TEXT NOT NULL,
    created_at      TIMESTAMPTZ NOT NULL DEFAULT now(),
    updated_at      TIMESTAMPTZ NOT NULL DEFAULT now()
);
CREATE INDEX claim_person_idx ON claim (person_id, claim_type, occurred_at DESC);
CREATE INDEX claim_org_idx ON claim (organization_id, claim_type, occurred_at DESC);
CREATE INDEX claim_cp_org_idx ON claim (counterpart_org_id) WHERE counterpart_org_id IS NOT NULL;
CREATE INDEX claim_type_time_idx ON claim (claim_type, occurred_at DESC);
CREATE INDEX claim_merged_idx ON claim (merged_into) WHERE merged_into IS NOT NULL;
CREATE INDEX claim_review_idx ON claim (review_status) WHERE review_status <> 'published';
COMMENT ON TABLE claim IS 'An atomic statement about a person and/or organization, never shown without evidence. Identity (fingerprint) is computed by the Loader from resolved entities and type-specific fields, so the same fact from several sources is one claim.';
COMMENT ON COLUMN claim.status IS 'reported = 1 source; corroborated = >= 2 independent sources; disputed = contradicting evidence; retracted.';
COMMENT ON COLUMN claim.sensitive IS 'Legal matters: UI and agents must show "Es gilt die Unschuldsvermutung" unless legal stage = convicted_final and not disputed.';
COMMENT ON COLUMN claim.merged_into IS 'Set when found to describe the same fact as another claim (corroboration); reads skip merged claims.';

CREATE TABLE claim_evidence (
    id              BIGSERIAL PRIMARY KEY,
    claim_id        BIGINT NOT NULL REFERENCES claim(id) ON DELETE CASCADE,
    document_id     BIGINT NOT NULL REFERENCES document(id) ON DELETE CASCADE,
    stance          TEXT NOT NULL DEFAULT 'supports' CHECK (stance IN ('supports', 'contradicts', 'mentions')),
    quote           TEXT,
    locator         TEXT NOT NULL DEFAULT '',
    extracted_by    TEXT NOT NULL,
    confidence      REAL NOT NULL DEFAULT 1.0,
    created_at      TIMESTAMPTZ NOT NULL DEFAULT now(),
    UNIQUE (claim_id, document_id, locator)
);
CREATE INDEX claim_evidence_doc_idx ON claim_evidence (document_id);

CREATE TABLE topic_assignment (
    topic_id        INT NOT NULL REFERENCES topic(id) ON DELETE CASCADE,
    claim_id        BIGINT NOT NULL REFERENCES claim(id) ON DELETE CASCADE,
    confidence      REAL NOT NULL DEFAULT 1.0,
    method          TEXT NOT NULL,
    PRIMARY KEY (topic_id, claim_id)
);
CREATE INDEX topic_assignment_claim_idx ON topic_assignment (claim_id);

-- ---------------------------------------------------------------------------
-- Facets
-- ---------------------------------------------------------------------------

CREATE TABLE role (
    claim_id        BIGINT PRIMARY KEY REFERENCES claim(id) ON DELETE CASCADE,
    person_id       INT NOT NULL REFERENCES person(id) ON DELETE CASCADE,
    organization_id INT REFERENCES organization(id),
    party_id        INT REFERENCES organization(id),
    role            TEXT NOT NULL,
    level           TEXT NOT NULL CHECK (level IN ('eu', 'federal', 'state', 'district', 'municipal', 'party', 'other')),
    region          TEXT,
    start_date      DATE,
    end_date        DATE
);
CREATE INDEX role_person_idx ON role (person_id);
CREATE INDEX role_party_idx ON role (party_id);
CREATE INDEX role_org_idx ON role (organization_id);
COMMENT ON TABLE role IS 'Mandates and functions. role = ''mp'' for members of a chamber (organization_id = chamber), otherwise the office title. end_date NULL = open.';

CREATE TABLE vote_event (
    id              BIGSERIAL PRIMARY KEY,
    event_key       TEXT NOT NULL UNIQUE,
    chamber_id      INT REFERENCES organization(id),
    item_ref        TEXT,
    item_title      TEXT NOT NULL,
    stage           TEXT,
    voted_on        DATE,
    result          TEXT CHECK (result IN ('accepted', 'rejected', 'unknown')),
    roll_call       BOOLEAN NOT NULL DEFAULT FALSE,
    url             TEXT,
    protocol_url    TEXT,
    raw_text        TEXT
);
CREATE INDEX vote_event_date_idx ON vote_event (voted_on DESC);
COMMENT ON TABLE vote_event IS 'A vote. Most Austrian votes are by standing up: only faction positions exist. roll_call = TRUE: individual positions exist.';

CREATE TABLE vote_position (
    claim_id        BIGINT PRIMARY KEY REFERENCES claim(id) ON DELETE CASCADE,
    vote_event_id   BIGINT NOT NULL REFERENCES vote_event(id) ON DELETE CASCADE,
    organization_id INT REFERENCES organization(id),
    person_id       INT REFERENCES person(id),
    party_id        INT REFERENCES organization(id),
    position        TEXT NOT NULL CHECK (position IN ('for', 'against', 'abstain', 'absent')),
    CHECK (organization_id IS NOT NULL OR person_id IS NOT NULL)
);
CREATE INDEX vote_position_event_idx ON vote_position (vote_event_id);
CREATE INDEX vote_position_person_idx ON vote_position (person_id) WHERE person_id IS NOT NULL;
COMMENT ON COLUMN vote_position.party_id IS 'For individual (roll-call) votes: the voter''s faction at the time.';

CREATE TABLE money_flow (
    claim_id        BIGINT PRIMARY KEY REFERENCES claim(id) ON DELETE CASCADE,
    kind            TEXT NOT NULL CHECK (kind IN ('donation', 'contract', 'government_ad', 'political_ad', 'subsidy',
                                                  'party_funding', 'party_ad', 'loan', 'side_income', 'sponsorship',
                                                  'other')),
    payer_name      TEXT,
    payer_org_id    INT REFERENCES organization(id),
    payer_person_id INT REFERENCES person(id),
    payee_name      TEXT,
    payee_org_id    INT REFERENCES organization(id),
    payee_person_id INT REFERENCES person(id),
    amount          NUMERIC(16, 2),
    amount_category TEXT,
    currency        TEXT NOT NULL DEFAULT 'EUR',
    paid_on         DATE,
    period          TEXT,
    reference       TEXT,
    category        TEXT
);
CREATE INDEX money_flow_payer_idx ON money_flow (payer_org_id);
CREATE INDEX money_flow_payee_idx ON money_flow (payee_org_id);
CREATE INDEX money_flow_kind_idx ON money_flow (kind, paid_on DESC);
COMMENT ON TABLE money_flow IS 'Donations (Rechnungshof lists and party reports), public party funding, sponsoring, ads in party media (party_ad), loans, public contracts (BVergG core data), government advertising (RTR), political ads (Google/Meta archives). category = CPV code, ad medium category etc.';

CREATE TABLE event (
    id              BIGSERIAL PRIMARY KEY,
    event_key       TEXT NOT NULL UNIQUE,
    kind            TEXT NOT NULL CHECK (kind IN ('parliament_session', 'committee', 'press_conference', 'tv_appearance',
                                                  'interview', 'party_event', 'public_event', 'foreign_trip', 'meeting',
                                                  'hearing', 'government_meeting', 'other')),
    title           TEXT NOT NULL,
    starts_at       TIMESTAMPTZ,
    ends_at         TIMESTAMPTZ,
    location_name   TEXT,
    country         TEXT,
    url             TEXT,
    description     TEXT,
    organization_id INT REFERENCES organization(id),
    source_id       INT REFERENCES source(id)
);
CREATE INDEX event_time_idx ON event (starts_at);
COMMENT ON TABLE event IS 'Something with a time and place (sessions, hearings, trips, meetings). Participation is a claim with an event_participant facet.';

CREATE TABLE event_participant (
    claim_id        BIGINT PRIMARY KEY REFERENCES claim(id) ON DELETE CASCADE,
    event_id        BIGINT NOT NULL REFERENCES event(id) ON DELETE CASCADE,
    person_id       INT REFERENCES person(id),
    organization_id INT REFERENCES organization(id),
    participation   TEXT NOT NULL DEFAULT 'participant'
);
CREATE INDEX event_participant_event_idx ON event_participant (event_id);

CREATE TABLE statement (
    claim_id        BIGINT PRIMARY KEY REFERENCES claim(id) ON DELETE CASCADE,
    person_id       INT REFERENCES person(id),
    context         TEXT NOT NULL CHECK (context IN ('parliament_speech', 'interview', 'tv', 'radio', 'social_post',
                                                     'press_release', 'event_speech', 'op_ed', 'testimony', 'other')),
    platform        TEXT,
    medium          TEXT,
    title           TEXT,
    said_at         TIMESTAMPTZ,
    excerpt         TEXT,
    url             TEXT NOT NULL
);
CREATE INDEX statement_person_idx ON statement (person_id, said_at DESC);

CREATE TABLE legal_proceeding (
    claim_id        BIGINT PRIMARY KEY REFERENCES claim(id) ON DELETE CASCADE,
    person_id       INT REFERENCES person(id),
    organization_id INT REFERENCES organization(id),
    stage           TEXT NOT NULL CHECK (stage IN ('allegation_reported', 'investigation', 'immunity_waiver_requested',
                                                   'immunity_waived', 'immunity_refused', 'indicted', 'trial',
                                                   'convicted_not_final', 'convicted_final', 'acquitted',
                                                   'proceedings_dropped', 'diversion')),
    offense         TEXT,
    authority       TEXT,
    case_ref        TEXT,
    stage_date      DATE
);
COMMENT ON COLUMN legal_proceeding.stage IS 'Only convicted_final (and not disputed) is a settled finding of guilt; all other stages carry "Es gilt die Unschuldsvermutung".';

-- ---------------------------------------------------------------------------
-- Retrieval (RAG)
-- ---------------------------------------------------------------------------

CREATE TABLE chunk (
    id              BIGSERIAL PRIMARY KEY,
    document_id     BIGINT REFERENCES document(id) ON DELETE CASCADE,
    claim_id        BIGINT REFERENCES claim(id) ON DELETE CASCADE,
    chunk_index     INT NOT NULL DEFAULT 0,
    text            TEXT NOT NULL,
    lang            TEXT DEFAULT 'de',
    tsv             TSVECTOR GENERATED ALWAYS AS (to_tsvector('german', text)) STORED,
    embedding       vector(1024),
    embedding_model TEXT,
    CHECK (document_id IS NOT NULL OR claim_id IS NOT NULL)
);
CREATE UNIQUE INDEX chunk_uniq ON chunk (COALESCE(document_id, 0), COALESCE(claim_id, 0), chunk_index);
CREATE INDEX chunk_tsv_idx ON chunk USING gin (tsv);
CREATE INDEX chunk_embedding_idx ON chunk USING hnsw (embedding vector_cosine_ops);
COMMENT ON TABLE chunk IS 'Text units for hybrid search (German FTS + bge-m3 embeddings); each links to a claim and/or document for citations.';

-- ---------------------------------------------------------------------------
-- Operations
-- ---------------------------------------------------------------------------

CREATE TABLE run (
    id              BIGSERIAL PRIMARY KEY,
    adapter         TEXT NOT NULL,
    mode            TEXT NOT NULL DEFAULT 'fetch',
    started_at      TIMESTAMPTZ NOT NULL DEFAULT now(),
    finished_at     TIMESTAMPTZ,
    status          TEXT NOT NULL DEFAULT 'running' CHECK (status IN ('running', 'ok', 'partial', 'error')),
    stats           JSONB NOT NULL DEFAULT '{}'::jsonb,
    error           TEXT
);
CREATE INDEX run_adapter_idx ON run (adapter, started_at DESC);

CREATE TABLE state (
    adapter         TEXT NOT NULL,
    key             TEXT NOT NULL,
    value           JSONB NOT NULL,
    updated_at      TIMESTAMPTZ NOT NULL DEFAULT now(),
    PRIMARY KEY (adapter, key)
);
COMMENT ON TABLE state IS 'Cursors of incremental adapters (last date, last id...).';

CREATE TABLE app_user (
    id              SERIAL PRIMARY KEY,
    external_subject TEXT NOT NULL UNIQUE,
    email           TEXT,
    locale          TEXT NOT NULL DEFAULT 'de',
    excluded_source_ids INT[] NOT NULL DEFAULT '{}',
    topic_ids       INT[] NOT NULL DEFAULT '{}',
    followed_person_ids INT[] NOT NULL DEFAULT '{}',
    created_at      TIMESTAMPTZ NOT NULL DEFAULT now()
);

-- ---------------------------------------------------------------------------
-- Views
-- ---------------------------------------------------------------------------

CREATE VIEW person_identifier AS
SELECT p.id AS person_id,
       max(i.value) FILTER (WHERE i.scheme = 'parlament') AS parlament_pad,
       max(i.value) FILTER (WHERE i.scheme = 'wikidata') AS wikidata_qid,
       max(i.value) FILTER (WHERE i.scheme = 'ep') AS ep_id
FROM person p LEFT JOIN identifier i ON i.entity_type = 'person' AND i.entity_id = p.id
GROUP BY p.id;
COMMENT ON VIEW person_identifier IS 'Convenience: the most common external ids of a person in columns.';

-- Activity status, derived from roles.
-- Parliament mandates/functions are authoritative (exact end dates) -> verified.
-- Other sources (mostly Wikidata) often lack end dates: an open-ended office of a living person
-- counts as current if it started within the last 15 years; flagged verified = false.
CREATE VIEW person_status AS
WITH current_roles AS (
    SELECT r.person_id, r.level, r.role, r.party_id, r.start_date, c.created_by
    FROM role r JOIN claim c ON c.id = r.claim_id
    WHERE c.merged_into IS NULL AND (r.end_date IS NULL OR r.end_date >= current_date)
      AND (c.created_by LIKE 'parlament%%' OR c.created_by LIKE 'europarl%%' OR r.end_date IS NOT NULL OR r.level = 'party'
           OR (r.start_date IS NOT NULL AND r.start_date >= current_date - interval '15 years'))
)
SELECT p.id AS person_id,
       CASE
           WHEN p.death_date IS NOT NULL THEN 'deceased'
           WHEN EXISTS (SELECT 1 FROM current_roles cr WHERE cr.person_id = p.id AND cr.level <> 'party') THEN 'active'
           WHEN EXISTS (SELECT 1 FROM current_roles cr WHERE cr.person_id = p.id AND cr.level = 'party') THEN 'party_member'
           ELSE 'former'
       END AS status,
       EXISTS (SELECT 1 FROM current_roles cr WHERE cr.person_id = p.id AND cr.level <> 'party'
               AND (cr.created_by LIKE 'parlament%%' OR cr.created_by LIKE 'europarl%%')) AS verified,
       (SELECT max(r.end_date) FROM role r JOIN claim c ON c.id = r.claim_id
        WHERE r.person_id = p.id AND c.merged_into IS NULL AND r.level <> 'party') AS last_office_end,
       (SELECT array_agg(DISTINCT cr.role) FROM current_roles cr WHERE cr.person_id = p.id AND cr.level <> 'party')
           AS current_roles,
       (SELECT array_agg(DISTINCT cr.party_id) FROM current_roles cr WHERE cr.person_id = p.id AND cr.party_id IS NOT NULL)
           AS current_party_ids
FROM person p;
COMMENT ON VIEW person_status IS 'active = current office (verified = confirmed by Parliament/EP data); party_member = only party membership; former; deceased.';

CREATE VIEW claim_with_evidence AS
SELECT c.id AS claim_id, c.claim_type, c.summary_de, c.summary_en, c.occurred_at, c.valid_from, c.valid_to,
       c.status, c.sensitive, p.full_name AS person_name, p.slug AS person_slug,
       o.name AS organization_name, o.slug AS organization_slug,
       jsonb_agg(jsonb_build_object('url', d.url, 'title', d.title, 'source', s.slug, 'source_name', s.name,
                                    'published_at', d.published_at, 'quote', e.quote, 'stance', e.stance)
                 ORDER BY d.published_at) AS evidence
FROM claim c
LEFT JOIN person p ON p.id = c.person_id
LEFT JOIN organization o ON o.id = c.organization_id
JOIN claim_evidence e ON e.claim_id = c.id
JOIN document d ON d.id = e.document_id
JOIN source s ON s.id = d.source_id
WHERE c.merged_into IS NULL AND c.review_status = 'published'
GROUP BY c.id, p.id, o.id;
COMMENT ON VIEW claim_with_evidence IS 'Published claims with all evidence links - the easiest entry point for agents.';

-- Read-only role for agents / BI. Password set by `politrack migrate` from POLITRACK_READER_PASSWORD.
DO $$
BEGIN
    IF NOT EXISTS (SELECT FROM pg_roles WHERE rolname = 'politrack_reader') THEN
        CREATE ROLE politrack_reader LOGIN;
    END IF;
END $$;
GRANT USAGE ON SCHEMA public TO politrack_reader;
GRANT SELECT ON ALL TABLES IN SCHEMA public TO politrack_reader;
REVOKE SELECT ON app_user FROM politrack_reader;
ALTER DEFAULT PRIVILEGES IN SCHEMA public GRANT SELECT ON TABLES TO politrack_reader;