References Tab¶

References page UI References tab — article cards with CELEX identifier, year, summary, and EUR-Lex link

Shows every EUR-Lex article used to produce the last answer. Each entry is a styled card with CELEX, year, article title, summary, and a direct EUR-Lex link.

File: app.py — with tab_refs:

Reference Card Fields¶

Field	Source file	Original field	Calculation	Business meaning
CELEX	`eurolex_client.py` SPARQL	`?celex` binding	Raw CELEX string from `doc_id`	Official EUR-Lex document ID (e.g. `32019R2144` = sector 3, year 2019, Regulation 2144)
Year	`tree_builder.py`	`tree["year"]`	`str(year)` or `"uploaded"` if year is 0	Year the document was created in EUR-Lex
Article	`tree_builder.py`	`node["title"]`	`f"Article {title}"`	Article number and title as parsed from document Markdown
Summary	`tree_builder.py` via Claude Haiku	`node["summary"]`	First 120 chars of Haiku-generated summary	One-sentence description of what the article covers
🔗 Open on EUR-Lex	`eurolex_client.py`	`doc["url"]`	`https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:{celex}`	Direct link to the full document

Data Flow¶

run_query() — app.py
  └─ selected nodes (from navigate_tree + rerank_nodes)
       └─ each node → reference dict:
            {CELEX, Article, Summary, URL, Year}
                 │
                 └─ st.session_state.references
                          │
                          └─ rendered in tab_refs as HTML cards

Download button: st.download_button() serialises references as json.dumps(refs, indent=2). Filename: eurolex_refs_{YYYYMMDD_HHMMSS}.json

Reference Dict Schema¶

{
  "CELEX":   "32019R2144",
  "Article": "Article 4 — General obligations of manufacturers",
  "Summary": "Requires manufacturers to ensure vehicles comply with type-approval requirements.",
  "URL":     "https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32019R2144",
  "Year":    "2019"
}