Aisystant Open

Find literal mentions of a known pattern: {"pattern_id":"SYSE.24"}. Read returned source context to establish each relation; a mention alone does not prove a dependency. Duplicate target IDs are marked ambiguous. Continue with offset = next_offset and expected_snapshot = snapshot, keeping pattern_id. Stop at next_offset null.

Check the current Library edition and update state when freshness, provenance or available scope matters. Call with {}: no arguments. Returns snapshot, source repositories/commits, corpus counts and last upstream check. Source integrity and freshness are separate; a failed refresh preserves the last working edition.

List headings in one publication. First get its id from list_publications; arguments require publication. Returns sections[].key, title and citation, plus snapshot. Copy key into read_section.section, not the heading title. To include prefaces/guides leave patterns_only false or omitted. Continue with offset = next_offset and expected_snapshot = snapshot. Stop when next_offset is null.

Discover publications, Suite References and guides. First call arguments: {} (or {"limit":20}). Returns publications[].id, title, path, suite and kind. Copy id into other tools' publication field; never infer it from a title or short name. Continue with offset = response.next_offset and expected_snapshot = response.snapshot, keeping filters unchanged. Stop when next_offset is null.

Read a complete pattern and its subsections as original Markdown in one call: {"pattern_id":"SYSE.24"}. Omit limit and offset for the whole text; the result has complete true and next_offset null. A bare ID follows retained source addresses; from a current search/ToC result also copy publication and snapshot into publication and expected_snapshot. For a subsection use read_section. Returns markdown, citation, offset, next_offset, total_characters and complete. Only clients needing smaller responses should set limit (1–60000), then continue with offset = next_offset and expected_snapshot = citation.snapshot until next_offset null.

Read a complete section, inclusive section range or publication as original Markdown in one call. Required: publication ID. Omit section to read the entire publication, e.g. {"publication":"<guide ID from list_publications>"}. Use known authored section IDs directly, including through: no ToC lookup is required. For a discovered section, copy results[i].publication and results[i].section from search and its top-level snapshot into expected_snapshot; a ToC sections[i].key also works. Optional through includes the ending heading and its subsections; it requires section. Temporary s_ keys require expected_snapshot even on the first read. Omit limit and offset for the whole selected text: complete true, next_offset null. Set limit (1–60000) only for clients needing smaller responses; continue with offset = next_offset and expected_snapshot = citation.snapshot, retaining publication/section/through. Returns markdown, citation, offset, next_offset, total_characters and complete.

Find source passages. Required argument q is the search string: {"q":"SYSE.24"} or {"q":"operating reliability","limit":5}. Meaningful words use AND; common English function words are skipped in plain queries. Exact IDs take priority. Returns results[].publication and results[].section plus top-level snapshot: copy these into read_section. Continue with offset = next_offset and expected_snapshot = snapshot, retaining q and filters. Read the source before relying on a snippet.