Pauhu

List all 14 Pauhu sectoral products with their launch state, coverage state, and the current access flag (`entitled`: true|false per row); access does not mean a product is sellable. IMPORTANT: `live_row_count` is a fast estimate of the shared source data, NOT the product's own filtered slice -- read `row_count_basis`/`row_count_note` beside it, and use `corpus_state` and `launch_state` as the coverage signals. A product may be entitled and still not be served yet. Data from this service is sold as the pauhu.eu bulk annual plan; nothing is sold per call here, and cited answers are at https://pauhu.ai. Search is never billed. Customer's AI client calls this once at session start to learn product and access state. Pass product_id for one product's full served row: what one delivered unit is per source (unit_definition), where its rows come from (provenance), its sectoral data space (sectoral_ds), licence, cadence, limitations.

Look up EU-canonical terms across 24 official EU languages in the EU's authoritative terminology record. Returns the authoritative EU term, domain, reliability rating, and equivalents in your target language. Accepts single terms or multi-word phrases. Used for professional translation and interpretation, using domain terminology sourced from the EU's own authoritative record.

Register one of your APIs or MCP servers as a connector, so your sourced monitors can watch it. Inputs: id (unique within your account), kind ('mcp' for an MCP server, 'https_json' for an HTTPS endpoint answering JSON with url plus text, paragraph_text or title, 'pauhu_sources' for the Pauhu sources your plan covers), endpoint_url, label, auth_kind ('bearer' with the name of your stored token, or 'none' for a public endpoint), and optionally mcp_tool_name (default 'search') and poll_fallback_secs. Returns the registered connector. Monitor runs over your connectors carry no per-run token charge.

List the connectors registered under your account.

Takes no parameters.

Delete a registered connector by id (tenant-scoped, you can only delete your own).

Register a sourced monitor and get back the registered template. Each match is delivered to you as a cited result: the source URL, the paragraph text and a verification status. Inputs: id, connector_ids (your registered connectors to watch), query, optional jurisdiction and language filters, when to push (every_match, new_since_last, or delta_at_least with a count) and where to push (a webhook URL with an optional secret reference, or sse for an open MCP session).

List the interest templates registered under your account.

Takes no parameters.

Delete a registered interest template by id (tenant-scoped).

Look up a text segment in Pauhu TM: aligned human-translated paragraphs from primary EU law sources. Returns the best exact or fuzzy match (gamma score 0.0-1.0) + EU terminology for the segment. Each match traces to a real paragraph URL in the original source document. Designed for CAT tool integration: pass a segment, receive the TM match + terms as a sourced CAT assist.

Search Pauhu's curated, sourced corpus. Returns sourced passages and per-match snippets, and every answer either traces to a source or says plainly when nothing matches.

Cross-reference one concept across EU and global authority id-spaces as aligned facets. Resolves the term to its concept anchor and emits a uniform list of facets: identifier CURIEs and language-tagged rendering literals ("label"@lang), each keyed by the shared concept, carrying one align verdict: exact (cited at its authority origin), frontier (no attested edge), or unresolved (transient outage, not 'no edge'). An exact identifier facet carries an IRI where its authority publishes one.

Search Pauhu's sourced index and get cited results as JSON. Input: q (query) and limit. Returns answer, results (id, title, url) and caller_verifies, a verification status: when true, open the cited text before relying on it. For one result's citation link, call source_fetch_fragment with the same q and limit and the result id. A citation shows where the text is quoted from, not that the source's claim is true or current; a general web publisher is one account, not established fact.

Get one cited result from source_search. Input: the q and limit you passed to source_search and a result id. Returns source_url, text_fragment_url (a link that opens the quoted text), caller_verifies (a verification status: when true, open the link before relying on the text) and bearing, a proof token to pass unchanged to source_verify_bearing, or null. If the results have changed since your search, the call says so; run source_search again. A citation shows where the text is quoted from, not that the source's claim is true; a general web publisher is one account, not established fact.

Lightweight liveness check for this resource. Returns { ok, status } with status ok, degraded or unavailable.

Takes no parameters.

Checks the bearing returned by source_fetch_fragment. Input: bearing, exactly as source_fetch_fragment returned it, and optionally the result id. Returns verified (true or false) and the id.

Reads one file of a public GitHub repository or npm package straight from the origin, at an exact identity: a branch or tag is resolved to its commit sha first, and an npm tarball must match the registry's sha512. Returns the requested lines with the commit (or version and integrity), the path, the git blob id of the whole file and a source URL, so the lines can be re-read and checked. Pages by line: give start_line/end_line; next_start_line names the next page. A file that is not UTF-8 text (a NUL byte in its first 8000 bytes, or invalid UTF-8) is never shown as text: the answer is binary:true with its reason, size and blob id. Read-only; nothing is indexed or stored. Every answer, an error included, carries a pauhu-library-receipt/2 in structuredContent.receipt and as a trailing "receipt: {...}" text line; library_verify checks a receipt.

Lists the files under a path of a public GitHub repository (at a commit sha) or npm package (at a version whose tarball matched the registry's sha512), with type, size and git blob id per file. Pages with offset/limit (at most 1000 entries and about 1 MB per page; truncated:true when the byte bound ended a page); total names the full count and next_offset the next page. Every answer, an error included, carries a pauhu-library-receipt/2 in structuredContent.receipt and as a trailing "receipt: {...}" text line; library_verify checks a receipt.

Finds a literal string (no regular expressions) in the text files of a public GitHub repository at a commit sha or an npm package version, and returns the matches with path, line number, line text and the file's git blob id. A hit is a proven match in the exact bytes named. Counts of files scanned, binary files (a NUL byte or invalid UTF-8; never searched as text) and files over the per-file memory bound are reported, so nothing is silently left out. Pages with offset/limit: at most 1000 matches and about 1 MB per page (truncated:true when the byte bound ended a page); next_offset names the next page. Every answer, an error included, carries a pauhu-library-receipt/2 in structuredContent.receipt and as a trailing "receipt: {...}" text line; library_verify checks a receipt.

Compares two exact identities of the same public GitHub repository (base and head refs, both resolved to commit shas) or npm package (base and head versions, both tarballs checked against the registry's sha512). Without path: the added, removed and changed files, decided by git blob id. With path: a unified diff of that file (Myers; minimal:false marks a large change answered as one replace hunk); identical blobs answer no hunks, a file that is not UTF-8 text answers binary:true with both blob ids and no hunks ("Binary files differ"), and a change region of more than 100000 lines is refused (diff_too_large). Pages with offset/limit (at most 1000 items and about 1 MB per page; truncated:true when the byte bound ended a page). A hunk larger than a page is split at line boundaries into consecutive hunks that apply in order; a single line larger than a page is refused (response_too_large) with both blob ids. Every answer, an error included, carries a pauhu-library-receipt/2 in structuredContent.receipt and as a trailing "receipt: {...}" text line; library_verify checks a receipt.

Lists what the code tools answer: for each, its id@version, the tool to call and the input it takes. Every code_* answer carries a receipt (format pauhu-library-receipt/2); pass it to library_verify to learn whether this service issued it and it is unchanged. Read-only; takes no input.

Takes no parameters.

Pass the receipt from any code_* answer and learn whether this service issued it and it is unchanged. Answers {verified, reason, id}: verified true or false; reason null, or one of bad_shape (not a receipt, or malformed), unknown_block (an id@version these tools do not answer), not_ours (not issued by this service), altered (changed after it was issued), unavailable (cannot be checked right now; retry later) or self_check_failed; id is the receipt id. With no input, runs the code tools' self-check and answers verified true or false. Read-only.