Every ComplyHat organization has a private compliance wiki — a structured knowledge graph of pages that back every citation in your reports and model cards. When your host agent generates a report, it can reference wiki pages as sources forDocumentation Index
Fetch the complete documentation index at: https://docs.complyhat.ai/llms.txt
Use this file to discover all available pages before exploring further.
[EXTRACTED] findings. You can also query the wiki directly to surface relevant policies, test configurations, and prior findings during an audit conversation. The wiki is per-tenant: your pages are never shared across organizations.
Write or update a page
Call thewiki tool with mode: "write" to create a new page or update an existing one. Pages are identified by a slug matching ^[a-z0-9][a-z0-9/_-]{0,254}$.
[[double-bracket-links]] in the body are automatically extracted and recorded as wiki links to other pages in your organization. The response includes the assigned id, the version number, a count of links_extracted, and a log_id from the immutable audit log.
To update a page and enforce optimistic concurrency, pass prev_version with the version number you last read:
409 VERSION_CONFLICT error.
Search the wiki
Thewiki tool supports two search modes: lexical (fast ILIKE on title and slug) and semantic (vector cosine similarity).
Lexical search
Usemode: "search" with search_mode: "index" for quick keyword lookups — good for finding pages you already know exist.
Semantic search
Usesearch_mode: "semantic" to find conceptually related pages. You must supply the embedding yourself — ComplyHat never calls an embedding model internally. Pass a 1536-dimensional vector generated by your host’s embedding function.
Embeddings are host-supplied. Your host agent generates the vector and passes it as the
embedding field — ComplyHat persists and queries it using pgvector but never generates embeddings itself. The same applies when writing a page: pass embedding on mode: "write" to make that page searchable semantically.501 SEMANTIC_UNAVAILABLE error. Fall back to search_mode: "index" for lexical search in that case.
Read and list pages
Read a single page by slug withmode: "read":
mode: "list". You can filter by kind:
entity, concept, comparison, summary, overview, finding, log.
Draft a page from a compliance finding
When a bias test or report surfaces a finding that warrants documentation, usemode: "draft_from_finding" to generate a structured wiki page draft. Pass the finding ID and the tool that produced it:
target_slug with kind: "finding" and a [INFERRED]-tagged body. Review and edit it before treating it as a signed-off record.
Next steps: See all
wiki tool modes — including link, lint, archive, and log_list — in the tool reference.