plenum/parliament.yaml

# Everything specific to one parliament, in one file. Nothing secret belongs here.
#
# This is the Swedish Riksdag configuration, shipped as the worked example. To
# target a different parliament, copy this file, change the values, and write an
# ingest adapter — see docs/PORTING.md.
#
# Override the path with PARLIAMENT_CONFIG=/etc/plenum/parliament.yaml so a
# deployment's own values live outside the repository and never appear in a diff.

schema_version: 1

parliament:
  id: se-riksdagen
  name: Sveriges riksdag            # in the language set below
  name_en: Swedish Parliament
  country: SE
  chamber: unicameral               # unicameral | lower | upper
  open_data_url: https://data.riksdagen.se
  data_start_year: 1993

language:
  # Postgres text-search configuration. Must match the database's app.fts_config
  # setting — the application refuses to start on a mismatch, because a silent
  # mismatch returns near-zero rows with no error.
  # See: SELECT cfgname FROM pg_ts_config;
  fts_config: swedish
  prompt_language: sv               # selects prompts/<lang>/
  name: svenska                     # the language's own name, for prompts
  name_en: Swedish                  # its English name, for English-language prompts
  locale: sv-SE
  # Characters that must survive verbatim into search queries. Transliterating
  # these silently breaks matching on most Swedish terms.
  preserve_characters: "åäö"
  months:
    januari: "01"
    februari: "02"
    mars: "03"
    april: "04"
    maj: "05"
    juni: "06"
    juli: "07"
    augusti: "08"
    september: "09"
    oktober: "10"
    november: "11"
    december: "12"

# Domain words injected into prompts, written in `prompt_language`. These exist
# because "motion", "yrkande" and "riksmöte" have no stable English equivalents;
# the database columns are neutral and the country's own word lives here.
vocabulary:
  member: riksdagsledamot
  member_plural: riksdagsledamöter
  chamber: kammaren
  speech: anförande
  speech_plural: anföranden
  document: motion
  document_plural: motioner
  proposal: yrkande
  proposal_plural: yrkanden
  committee: utskott
  constituency: valkrets
  session: riksmöte

# Display order is list order. `active: false` keeps historical parties
# renderable without offering them as a filter.
parties:
  - {code: S,   name: Socialdemokraterna,    color: "#E8112d", active: true}
  - {code: M,   name: Moderaterna,           color: "#52BDEC", active: true}
  - {code: SD,  name: Sverigedemokraterna,   color: "#DDDD00", active: true}
  - {code: C,   name: Centerpartiet,         color: "#009933", active: true}
  - {code: V,   name: Vänsterpartiet,        color: "#b51a0e", active: true}
  - {code: KD,  name: Kristdemokraterna,     color: "#000077", active: true}
  - {code: MP,  name: Miljöpartiet de gröna, color: "#83CF39", active: true}
  - {code: L,   name: Liberalerna,           color: "#006AB3", active: true}
  - {code: NYD, name: Ny demokrati,          color: "#ffff2b", active: false}

party_defaults:
  # Independents and unknowns.
  unknown_color: "#9aa5b8"
  # How a party code appears in rendered text, e.g. "Anna Andersson (MP)".
  # Used to detect and verify attributions in generated answers.
  code_pattern: "[A-ZÅÄÖ]{1,3}"

# Keys are the values stored in talks.kammaraktivitet. Titles and descriptions
# are shown in the UI and are in `prompt_language`.
activity_types:
  kam-vo: {title: Beslut,                   description: "Riksdagen fattar det formella beslutet i ärendet."}
  bet:    {title: Debatt om beslut,         description: "Ledamöterna debatterar ett förslag innan beslut fattas."}
  kam-fs: {title: Frågestund,               description: "Ministrar svarar direkt på frågor från ledamöter."}
  kam-ar: {title: Information från regeringen, description: "Regeringen informerar riksdagen i en aktuell fråga."}
  ip:     {title: Interpellationsdebatt,    description: "Debatt om en skriftlig fråga till en minister."}
  kam-sf: {title: Statsministerns frågestund, description: "Statsministern svarar på frågor från ledamöter."}
  sam-ou: {title: Öppen utfrågning,         description: "Utskottet frågar ut experter eller ansvariga personer offentligt."}
  kam-ad: {title: Aktuell debatt,           description: "Snabb debatt om en aktuell politisk fråga."}
  kam-al: {title: Allmänpolitisk debatt,    description: "Ledamöterna debatterar valfria politiska ämnen."}
  kam-bu: {title: Budgetdebatt,             description: "Debatt om regeringens budgetförslag."}
  kam-bp: {title: Bordläggning,             description: "Ett ärende anmäls i kammaren inför fortsatt behandling."}
  kam-pd: {title: Partiledardebatt,         description: "Partiledarna debatterar politikens stora frågor."}
  kam-dv: {title: Debatt med anledning av vårpropositionen, description: "Debatt om regeringens ekonomiska vårproposition."}
  sam-se: {title: Öppet seminarium,         description: "Öppet seminarium eller samtal om ett särskilt tema."}
  kam-ud: {title: Utrikespolitisk debatt,   description: "Debatt om Sveriges utrikespolitik."}

# Glosses for values stored in the data. The values themselves are never
# translated — only explained, so prompts can reason about them.
document_subtypes:
  Enskild motion: "Motion från en eller ett fåtal ledamöter"
  Kommittémotion: "Motion från ett partis utskottsgrupp"
  Partimotion: "Motion från ett helt parti, ofta undertecknad av partiledaren"

decisions:
  Bifall: {label: Bifall, meaning: approved}
  Avslag: {label: Avslag, meaning: rejected}

# Annual parliamentary session. Deliberately not called "term": the European
# Parliament uses that for its five-year cycle, so it would be ambiguous.
sessions:
  label_format: "{start}/{end_short}"   # 2022 -> "2022/23"
  starts_month: 9
  first_year: 1990

# Identifier shapes, templated into prompts and tool descriptions so the model
# can recognise and construct them without them being hardcoded in Python.
ids:
  person_id:
    example: "0448485371626"
    pattern: "^[0-9]{10,13}$"
    description: "Riksdagen intressent-id, a numeric string"
  speech_id:
    example: "GH09116-16"
    pattern: "^[A-ZÅÄÖ0-9]+-[0-9]+$"
    description: "{source_doc_id}-{sequence}"
  doc_id:
    example: "HD02846"
    pattern: "^[A-ZÅÄÖ0-9]+$"
    description: "Document id from the open-data portal"
  debate_id:
    example: "2021-06-17:42"
    pattern: '^\d{4}-\d{2}-\d{2}:\d+$'
    description: "{date}:{index within that day}"

urls:
  person_photo: "https://data.riksdagen.se/filarkiv/bilder/ledamot/{person_id}_192.jpg"
  person_page: "https://www.riksdagen.se/sv/ledamoter-partier/ledamot/_{person_id}"
  document_page: "https://www.riksdagen.se/sv/dokument-och-lagar/dokument/_{doc_id}"

# Consumed by the ingest CLI. `adapter` names the module that knows this
# source's JSON shape; everything else is data.
sources:
  adapter: ingest.adapters.riksdagen
  speeches:
    kind: zip-dataset
    url_template: "https://data.riksdagen.se/dataset/anforande/anforande-{range}.json.zip"
    dest_dir: talks
  documents:
    kind: zip-dataset
    url_template: "https://data.riksdagen.se/dataset/dokument/mot-{range}.json.zip"
    ranges: ["1990-1997", "1998-2001", "2002-2005", "2006-2009",
             "2010-2013", "2014-2017", "2018-2021", "2022-2025"]
    dest_dir: motioner
  people:
    kind: json
    url: "https://data.riksdagen.se/personlista/?utformat=json"
    dest_dir: personer

embeddings:
  model: qwen3-embedding             # LLM_MODEL_EMBEDDING overrides
  # Must equal the vector(N) width of the embedding columns. Checked at startup
  # against the actual column type, because a mismatch fails deep inside pgvector
  # with an error that says nothing about configuration.
  dimension: 384
  base_url_env: EMBEDDING_BASE_URL
  chunk_chars: 500

# Served to the frontend via GET /api/meta.
site:
  # Deployment identity, not country data. Upstream ships neutral values; a
  # deployment overrides them in its own PARLIAMENT_CONFIG file so its branding
  # never appears in a diff against upstream.
  title: plenum
  tagline: "Sök i riksdagens anföranden och motioner"
  # Markdown, so a deployment can rewrite its own copy without touching source.
  # Paths are relative to CONTENT_DIR (default: ./content).
  explainer_file: sv/explainer.md
  limit_warning_file: sv/limit_warning.md
  # Upstream ships these empty. A deployment fills them in via its own
  # PARLIAMENT_CONFIG file, which keeps personal contact details out of the repo.
  guide_file: sv/user-guide.md
  contact:
    email: null
    url: null