Amok

Exact database rows behind corpus-quality gaps (incomplete fetches, non-binding-only text, unresolved relations, etc.) so an agent can see what is missing. Evidence about completeness, not legal content.

quality_gaps

Live corpus coverage: document/chunk/embedding counts, relation coverage, and known data gaps. Call this to gauge how complete the evidence is before relying on it.

corpus_status

Read first. Explains what amok covers and how to use its evidence tools (search → document) to answer a Cambodian banking/technology regulation question with exact citations — no local files or extra prompts needed.

guide

Open one legal document by id or doc_number (document reference): full provision text (reassembled Article/sub-article), validity periods, confirmed relations, verbatim incoming amendments, the official source link(s), and data gaps. Use it to read complete provisions when search returns fragments. It also returns incoming_amendments: verbatim clauses from documents that amend/replace this one (text + position) — read these to judge which provisions are still current. Pass include to fetch only the sections you need — include=['chunks'] with a citation filter (e.g. 'Article 26') is the cheapest way to read one provision. Returns content + source links only — never file downloads.

document

Search Cambodian banking & financial-technology regulation and return exact, citable evidence — ranked provisions (Article/Clause/sub-items) with their source document, validity status, confirmed relations, the official source link, and explicit gaps. No LLM synthesis: you get the source evidence and decide the answer. Use this whenever the question touches Cambodian banking/finance law or regulation, especially digital/technology topics: banking & financial institutions, payment systems (Bakong, KHQR), AML/CFT, microfinance, financial leasing, cryptoassets, securities, IT risk management, and technology operations. Query in English. detail levels: compact (discovery — metadata + snippet + cite, cheapest) | standard (default — adds relations + related_hits; no inline article text) | full (inlines every hit's whole Article — largest; prefer the document tool to read one provision).

search