Midlyr

Browse Midlyr's compliance-source catalog by keyword. `query` is trigram substring match against title + description — NOT semantic search; use `queryRegulations` for natural-language questions. To read a document, call `readRegulatoryDocument` (metadata + TOC) or `readRegulatoryContent` (body text).

Detailed compliance review of submitted text — heavier than `riskAssessment` (use that for a quick numeric score only). Defaults to inline wait up to ~9 minutes; set `waitForResult=false` to start the job and poll `getJob` instead.

Persist a compliance event as a ticket for the current team — fire-and-forget; downstream classification is async. Use `externalRef` to deduplicate retries. Use `complianceScreening` or `riskAssessment` instead when you need an analysis result inline.

Poll for the status of an asynchronous analysis job by id (from `complianceScreening` or `listJobs`). Only `screen_analysis` and `risk_assessment` jobs are visible — unknown ids, ids owned by another team, or jobs of other types are reported as not found rather than raising.

List recent analysis jobs for the current team, newest first — use `getJob` for full results. Only `screen_analysis` and `risk_assessment` are exposed. Defaults to the last 30 days when `start` and `end` are both omitted.

Semantic search across Midlyr's regulatory corpus for natural-language topical questions — use when the user has not named a specific document. Pure vector retrieval, no LLM. Use `browseRegulatoryLibrary` instead when you already have a document name or id.

Fetch a byte-range slice of a compliance document's body text. Pair with `readRegulatoryDocument` to get `totalBytes` and TOC before paging. To continue through a long document, set the next call's `offset` to the previous response's `content.endOffset` (NOT `offset + limit` — the read may return fewer bytes than requested) and stop when `content.hasMore` is `false`.

Fetch one compliance document's metadata, total text size, and table of contents. Does NOT return body text — use `readRegulatoryContent` for that.

Fast compliance risk score for submitted text — synchronous, typically under 20 s. Lighter-weight sibling of `complianceScreening`; use this for triage or live-feedback UIs. Use `complianceScreening` when the user needs violation details or regulation citations.