Two desk statuses
Core desks have an explicit profession-specific indicator configuration. Preview desks preserve the profession page, live global context and role-specific orientation while specialized KPI coverage is still being built. A Preview desk is never presented as though inherited metrics from another profession are its own.
Current public live feeds
CoinGecko — digital assets
The public market surfaces request cryptocurrency market data from CoinGecko. Where a panel is labeled live, the page fetches the current response and displays a failure state rather than substituting a believable static value after a feed error.
Frankfurter / ECB — foreign exchange
FX panels use Frankfurter, which distributes European Central Bank reference-rate data. ECB reference rates update on business days, so “live” means a fresh retrieval of the latest available reference-rate dataset — not tick-by-tick interbank pricing.
Hacker News — attention pulse
Pulse uses the public Hacker News API for its current attention feed. The analyst layer groups and summarizes the retrieved stories; it does not claim Hacker News is a comprehensive financial-news wire.
Reference indicators
Profession KPI tiles inherited from the historical build are reference baselines unless and until a secure production source integration refreshes them. Reference values can still be useful for orientation, but they must not be confused with a current quote.
For paid launch, each material desk signal needs a documented source, refresh cadence, stale threshold, role-specific explanation and safe failure state.
The Aria analyst engine
The current public analyst brief is deterministic. It reads the dataset fetched by the page, calculates breadth, dispersion, outliers and trend posture, then composes a structured note from those values. The same data snapshot produces the same analytical logic.
- Breadth: share of tracked assets advancing over selected periods.
- Movers and outliers: largest moves and two-standard-deviation exceptions.
- Risk regime: relationship between Bitcoin, Ethereum and the broader basket.
- Trend posture: daily versus weekly breadth.
- FX backdrop: context from the latest available ECB reference-rate dataset.
Role-based intelligence boundaries
ValixData is designed to enforce professional access levels at the data, report and AI-context layers. A lower access level should not be able to request, infer or reconstruct a higher level merely by changing a URL, browser state or prompt. Production entitlement checks belong server-side and should govern dataset scope, historical depth, cross-desk joins, report templates, exports, alerting and the context packet supplied to AI.
- Professional: focused operational facts, core signals and concise interpretation.
- Manager: broader comparisons, exceptions, trend context and limited adjacent-function visibility.
- Executive: cross-functional synthesis, scenario context and strategic exposures.
- Enterprise / C-suite: governed multi-desk, capital, risk and organization-level synthesis.
Human judgment is the control layer
ValixData may summarize facts, explain context, compare scenarios, identify anomalies and provide clearly qualified directional analysis. It is not designed to make consequential decisions on behalf of the user. It does not autonomously execute transactions, issue personalized buy/sell commands, approve expenditures, direct personnel actions or substitute for the accountable professional. The purpose of the intelligence layer is to improve human judgment, not remove it.
Aria + Nova cost architecture
ValixData uses a retrieval-first hierarchy: reuse approved desk knowledge and cached briefs first, calculate deterministic statistics second, and invoke managed generative AI only when bounded synthesis adds material value. Managed calls must run server-side with role entitlements, caching, deduplication, quotas, rate limits and spend caps.
Nova · reusable desk briefing
Nova is the scheduled briefing layer. The public preview can deterministically synthesize a recent approved-news window for each desk. Production Nova is designed to refresh once per desk and entitlement cohort, or when a material event warrants it, then cache the result for all entitled readers. It is not a per-user overnight model job.
Aria · grounded desk assistance
Aria is the explanatory layer. It starts with the user's entitled ValixData desk context, definitions, current signals, provenance and help content. Deterministic answers and retrieval should satisfy common questions without a model call; managed AI is reserved for bounded questions that require synthesis. No customer-supplied AI keys are permitted.
Data-integrity rules
- Never label a reference/static number as live.
- Never describe an unrelated inherited KPI set as profession-specific.
- Name the source and cadence for material paid indicators.
- Show stale/error states rather than inventing substitute values.
- Keep secret-backed credentials server-side.
- AI may interpret supplied facts; it may not invent new figures.
Refresh cadence
- Terminal & Markets: public market requests refresh on the page cadence shown in the interface.
- FX: retrieval can refresh frequently, but the underlying ECB reference dataset is business-day based.
- Pulse: refreshed less aggressively because ranking moves on a slower timescale.
- Profession KPIs: reference until an explicit production integration says otherwise.
What ValixData does not do
- It does not route orders or execute trades.
- It does not provide personalized investment advice.
- It does not ask customers for external AI-service API keys.
- It does not hide Preview status behind marketing language.
Self-service operations
The production service is intended to be self-administered. Account creation, billing, plan changes, seat administration, role assignment, desk configuration, watchlists, alert preferences, password recovery and cancellation should be available without routine human support. A documentation-grounded help assistant handles ordinary product questions; human support is an exception path.
Production architecture
The public site can remain lightweight and crawlable. Paid accounts, Merchant-of-Record billing, transactional email, managed AI and secret-backed data belong behind a small secure backend/serverless layer with rate limits, logs, spend controls and source-health monitoring.