alize_ask
ChatGPTRun a full legal research pass over Alizé's corpus for one jurisdiction and return a written answer with numbered citations to the specific pages and lines that support each point. Unlike alize_search, this plans its own retrieval across the jurisdiction's principal sources, reads what it finds and synthesises an answer. Use it for questions that need reasoning across several instruments — how a rule applies, what the courts have held, whether an obligation exists — rather than for finding a document you could name. Takes roughly 15-45 seconds and is rate-limited. When you only need to locate documents, alize_search is faster and cheaper.
alize_flat_search
ChatGPTList documents in one document type — title, date, identifiers, official URL and doc_id — with none of their text. Returns up to 25 documents per page against 5 for alize_search, because omitting page excerpts makes each result a few lines instead of several pages. Use it to survey what exists before deciding what to read: "what has the CJEU decided on this since 2020", "which circulars did the regulator publish last year". text_query is optional. Omit it to enumerate the collection by date, newest first — with from/to that answers "what was published in this window" without needing a subject at all. Every row carries the doc_type and doc_id that read_page, alize_search_in_doc and alize_regex_in_doc take. Reach for alize_search instead when you want the text of the best matches straight away.
alize_regex_in_doc
ChatGPTSearch one document for an exact pattern and return every match with the surrounding words and the page it is on. Returns up to 10 matches per call, each with about 10 words either side. Use this when only the literal string will do: "Article 17", a case or CELEX number, a defined term you need every occurrence of, a cross-reference. It answers "does this document ever say X, and where" — which semantic search cannot, because it will happily return something that means the same thing and never says it. Patterns are POSIX extended regular expressions. Character classes, quantifiers, alternation, anchors and case-insensitive mode "(?i)" all work; lookahead "(?=…)", lookbehind "(?<=…)" and backreferences do not exist and will be refused. Document text keeps its original language and accents, so match what is actually written — "société" not "societe". When you want meaning rather than an exact string, use alize_search_in_doc.
alize_search
ChatGPTSearch Alizé's curated corpus of official legal documents within one document type, combining keyword and semantic retrieval with relevance reranking. Returns the 5 best-matching documents per page, each with its citation metadata, official source URL, and the text of its first page, its two most query-relevant pages and its last page. Reach for this whenever an answer turns on what a statute, regulation or court decision actually says in Luxembourg, Belgium, Spain, France, the EU, the ECHR or Benelux. The corpus is ingested from official sources and kept current, so it is more reliable than recalling legal text from memory — and it returns a citable URL for every result. Write the query in English if you like: it is translated into the collection's language automatically. Document text comes back in its original language, which is the authoritative version. Requires a doc_type from list_types. Use read_page to read any other page of a result. If the results disappoint, do not rephrase and search again. Retrieval here is semantic, so a reworded query is the same query and returns practically the same documents — repeating it costs a call and changes nothing. What does change the results: a different doc_type (the most common reason for a thin result set is searching the wrong collection), a from/to range, or alize_flat_search to see what the collection actually holds. To go deeper into a document you already have, use alize_search_in_doc or read_page rather than searching the whole corpus again.
alize_search_in_doc
ChatGPTFind the passages inside one document that address a question, using the same keyword-plus-semantic retrieval as alize_search but scoped to a single document. Returns the 2 best-matching passages per call, each with the page either side of it so the passage reads as continuous text rather than a fragment. This is the right tool once you have a document and need the part of it that matters — a specific article in a long code, the paragraph of a judgment that states the holding, the clause a directive devotes to one obligation. Far cheaper than paging through with read_page, and it finds passages that use different words than your question. It matches on meaning, so you do not need the document's exact wording or language. When you do need an exact string — an article number, a reference, a defined term — use alize_regex_in_doc.
list_families
ChatGPTList the kinds of legal document Alizé holds for one jurisdiction — legislation, case law, regulatory-body publications, parliamentary files, doctrine. Call after list_jurisdictions; the returned family values feed list_types.
list_jurisdictions
ChatGPTList the legal jurisdictions Alizé covers, such as Luxembourg, Belgium, Spain, France, the European Union, the ECHR and Benelux. Start here when you do not already know whether Alizé holds material for a jurisdiction. Instant and free; the returned jurisdiction values feed list_families and alize_ask.
list_types
ChatGPTList the specific searchable document types in one jurisdiction and family, each with a description, its category and the language its documents are written in. The doc_type values returned here are what alize_search and read_page take — there is no other way to obtain a valid one.
read_page
ChatGPTReturn the full text of one page of a document found by alize_search. Use it when a search excerpt was truncated, when you need the pages around a match, or when a citation in an alize_ask answer points at a page you want to read in full. Requires the doc_type and doc_id exactly as they appeared in the result you are following up.
alize_ask
Claudealize_flat_search
Claudealize_regex_in_doc
Claudealize_search
Claudealize_search_in_doc
Claudelist_families
Claudelist_jurisdictions
Claudelist_types
Clauderead_page
Claude
