Published vocabulary
The application-domain vocabulary DSAIL rulesets are labeled with: the closed level-one list, every level-two term with its provenance, the promotion rule and K, and the version history — rendered live from the service.
Every ruleset may carry an application-domain label produced by the customer's own model and stored as ids from this vocabulary — never as text. This page is the human-readable rendering of the machine-readable document at <https://agents.jaxon.ai/v1/vocabulary>, which is the authority; the JSON is what a program should read.
How a term gets in. Level one is a closed list, revised only by a versioned release. Level two grows by promotion: a term your model proposes is matched against the current list and, when it matches nothing, counted. When at least K distinct external accounts have independently proposed the same normalised term it is promoted into this vocabulary with provenance promoted; Jaxon's own accounts never count; proposals that do not reach K within the TTL are deleted. K, the TTL and the exact rule text are in the live document under promotion.
What the live document adds to this page. Every term promoted since the shipped file, the current K and TTL, the counts, and version_history. The HTML twin of this page fetches and renders it; the markdown twin names it so a model fetches the JSON.
Current version
Shipped file: v1, published 2026-09-06.
Application-domain vocabulary for DSAIL rulesets (TJP-642). Level one is a closed, slow-moving list. Level two is an open vocabulary: the terms below are a sparse cold-start seed drawn from design-partner and sales conversations, and every further term enters only by promotion (K distinct external accounts proposing it independently). Events and stored labels carry ids, never strings; this document is where an id gets its name.
Level one (closed)
| Id | Name | Covers |
|---|---|---|
finance | Corporate finance and spend | Expense and reimbursement, spending accounts and approval limits, treasury, accounting, tax. |
consumer_finance | Consumer finance and lending | Loan and mortgage underwriting, credit decisions, retail banking, collections, fair-lending rules. |
capital_markets | Capital markets and trading | Trading desks, hedge funds and asset managers: trading limits, position and exposure rules, asset pricing and mispricing screens, fixed income and distressed credit, fund mandates. |
insurance | Insurance and reinsurance | Underwriting guidelines, claims adjudication, treaty and facultative reinsurance terms, coverage eligibility. |
regulatory_compliance | Regulatory compliance | Rules that restate an external regulator's requirements: FINRA, SEC, banking and market regulation, AML/KYC, reporting obligations. |
hr_workforce | HR and workforce | Hiring, leave, conduct, compensation, benefits, contractor and workforce policy. |
security_access | Security and access | Identity, access control, data handling, incident response and security controls. |
security_classification | Security classification and release | Classified or controlled information: classification guides, marking, need-to-know, release and disclosure restrictions, export control. A category only — controlled content itself does not belong on a public model or service. |
healthcare | Healthcare | Clinical, coverage, eligibility, privacy and care-delivery policy. |
legal_contracts | Legal and contracts | Contract terms, obligations, legal review and approval. Regulator-driven rules go under regulatory_compliance. |
procurement | Procurement | Vendor selection, purchasing authority, sourcing and supplier policy. |
engineering_release | Engineering and release | Change management, release gates, code review, deployment and operational readiness. |
content_safety | Content and safety | Acceptable use, moderation, publishing standards and safety review. |
other | Other | Nothing above fits. The share of the corpus landing here is the trigger for revising this list. |
The share of rulesets landing in other is the trigger for revising this list.
Level two, as seeded in v1
Seeded terms are a cold-start guess drawn from design-partner and sales conversations, marked seeded until K distinct external accounts have proposed them independently, at which point their provenance flips to promoted. Terms promoted after this file shipped appear only in the live document.
| Id | Term | Aliases | Provenance at seed |
|---|---|---|---|
l2-0001 | expense reimbursement | expense policy, travel and expense, reimbursement policy | seeded |
l2-0002 | loan underwriting | mortgage underwriting, credit underwriting | seeded |
l2-0003 | spending authority | approval limits, signing authority, spending accounts | seeded |
l2-0004 | leave and time off | pto policy, leave policy | seeded |
l2-0005 | hiring eligibility | candidate eligibility | seeded |
l2-0006 | access control | access management, least privilege | seeded |
l2-0007 | data classification | data handling | seeded |
l2-0008 | incident response | security incident | seeded |
l2-0009 | coverage eligibility | benefit eligibility | seeded |
l2-0010 | prior authorization | preauthorization | seeded |
l2-0011 | contract review | contract approval | seeded |
l2-0012 | regulatory compliance | compliance review | seeded |
l2-0013 | vendor onboarding | supplier onboarding | seeded |
l2-0014 | purchase approval | purchasing approval, purchase order approval | seeded |
l2-0015 | release readiness | release gate, deployment readiness | seeded |
l2-0016 | change management | change approval | seeded |
l2-0017 | acceptable use | acceptable use policy | seeded |
l2-0018 | content moderation | moderation policy | seeded |
l2-0019 | tax treatment | tax policy, tax compliance | seeded |
l2-0020 | treasury and cash management | treasury policy, cash management | seeded |
l2-0021 | credit decisioning | credit decision, credit approval, consumer credit | seeded |
l2-0022 | fair lending | fair lending compliance, ecoa | seeded |
l2-0023 | trading limits | position limits, risk limits, exposure limits | seeded |
l2-0024 | mispriced assets | whale shit, asset mispricing, bond mispricing, low grade bond screening, distressed credit screening | seeded |
l2-0025 | fund mandate compliance | investment mandate, mandate restrictions, investment guidelines | seeded |
l2-0026 | pre trade compliance | pre-trade checks, order compliance, restricted list | seeded |
l2-0027 | insurance underwriting | underwriting guidelines, policy underwriting | seeded |
l2-0028 | claims adjudication | claims handling, claims policy | seeded |
l2-0029 | reinsurance treaty terms | reinsurance, treaty compliance, facultative reinsurance | seeded |
l2-0030 | finra compliance | finra rules, finra | seeded |
l2-0031 | sec regulation | sec rules, sec compliance, securities regulation | seeded |
l2-0032 | aml and kyc | anti money laundering, know your customer, kyc, aml | seeded |
l2-0033 | classification marking | security classification guide, scg, classification guide, marking rules | seeded |
l2-0034 | release and disclosure | public release review, disclosure restrictions, need to know | seeded |
l2-0035 | export control | itar, ear, export compliance | seeded |
Version history
A version bump is a release that changes what stored ids mean (a level-one revision; renaming, merging, splitting or retiring level-two terms) and ships with a mapping from the prior version. Adding a term never needs one. The live document's version_history lists every version the service knows.
| Version | Published | Mapping from prior |
|---|---|---|
v1 | 2026-09-06 | none (first version) |
Live document
GET https://agents.jaxon.ai/v1/vocabulary — JSON, no credential required, readable from this site's origin. The terms version that governs a tier is reported beside the vocabulary version by dsail_get_account_status; the current published terms are version 2026-09-14.
Related: Data handling, Privacy, Terms of service.
Live vocabulary
Loading the live document from https://agents.jaxon.ai/v1/vocabulary…