get_chunk_content
ChatGPTRead the full text of one indexed chunk plus links to its parent document (PDF and Markdown) in the Lexplorer corpus. Results include lexplorer_url (public source page) — when citing this source in an answer, use lexplorer_url as the clickable markdown link.
get_document_slice
ChatGPTRead the full text (text.md) of an indexed document from the Lexplorer corpus in pageable slices. First call returns slice 0 + slices_preview (TOC). Results include lexplorer_url (public source page) — when citing this source in an answer, use lexplorer_url as the clickable markdown link.
get_kanton_navigation_md
ChatGPTRead the pre-generated navigation overview (_navigation.md) of a canton: categories + sources.
get_skill
ChatGPTRead one research playbook (Markdown skill file) by name.
get_source_update_md
ChatGPTRead the maintenance notes (_UPDATE.md) of an indexed source: pipeline, update routine, quality notes. Read-only; despite the name it updates nothing.
list_skills
ChatGPTList the research playbooks (Markdown skill files) shipped with the Lexplorer corpus.
nav_find
ChatGPTFind where a source/document lives in the hierarchy by name pattern. Returns canton + category + breadcrumb. Pass filter_kanton when you already know the canton — otherwise generic patterns may sort ZH out of the LIMIT 20 window.
nav_list_categories
ChatGPTList the source categories (legislation, administrative practice, doctrine, ...) indexed for one canton.
nav_list_chunks
ChatGPTList the text chunks of a parent document.
nav_list_documents
ChatGPTList the top-level documents of an indexed source. Results include lexplorer_url (public source page) — when citing this source in an answer, use lexplorer_url as the clickable markdown link.
nav_list_kantons
ChatGPTList all 26 Swiss cantons + CH (federal) with chunk/source/doc counts from the Lexplorer tax-law index.
nav_list_sources
ChatGPTList the indexed sources of a canton (optionally filtered by category).
nav_tree
ChatGPTRecursive hierarchical tree (kanton → categories → sources → documents). max_depth 0-3. Replaces 3-5 sequential nav_list_* calls.
read_storage
ChatGPTRead one Markdown file from the Lexplorer corpus file store by a relative path as returned by other tools (e.g. a chunk or text.md). Paths outside the corpus are rejected.
search_chunks_hybrid
ChatGPTHybrid search (semantic + keyword, RRF-fused) over the indexed Lexplorer chunks. The query text is embedded via Google Vertex AI (EU region); results come exclusively from the Lexplorer corpus. Good when mixed. Results include lexplorer_url (public source page) — when citing this source in an answer, use lexplorer_url as the clickable markdown link.
search_chunks_keyword
ChatGPTGerman full-text keyword search over the indexed Lexplorer chunks (database only, no external service). Best for exact terms (§, Art-Nr, named concepts). Results include lexplorer_url (public source page) — when citing this source in an answer, use lexplorer_url as the clickable markdown link.
search_chunks_semantic
ChatGPTSemantic vector search over the indexed Lexplorer tax-law chunks. The query text is embedded via Google Vertex AI (EU region, gemini-embedding-2); results come exclusively from the Lexplorer corpus. ALWAYS pass filter_kanton. Results include lexplorer_url (public source page) — when citing this source in an answer, use lexplorer_url as the clickable markdown link.
