Real-time inference
The models process quote flows as they happen, without batch processing. A change in trend on a pair is detected within seconds, not at the end of the day.
Lueur Fondaru combines real-time inference and predictive models to transform massive data streams into actionable signals, without manual intervention on each asset.
Each module addresses a specific part of the market flow. Together, they shorten the time between a change in market regime and your action.
The models process quote flows as they happen, without batch processing. A change in trend on a pair is detected within seconds, not at the end of the day.
Momentum indicators and inter-market correlations are continuously recalculated to estimate the probability of short-term movements on each instrument tracked.
Volatility clusters are identified automatically in order to adjust alert thresholds according to the current market regime, rather than using fixed parameters.
The Lueur architecture aggregates more than 500 market sources — order books, quote feeds and derived indicators — then applies a series of statistical filters to rule out variations without directional scope.
The result transmitted to the terminal is not a raw flow: it is a selection of signals deemed statistically significant, with their associated confidence level.
The goal is not to add data, but to reduce data that you have to process yourself. Three functions concentrate most of this work.
Rather than manually monitoring dozens of screens, the engine scans all the monitored instruments and only brings up those whose technical conditions match your criteria. This frees up your mental bandwidth for analysis and the final decision.
News feeds and market discussions are processed to extract a dominant orientation, weighted by the reliability of the source. This reading complements the technical signals, without replacing them.
You set the thresholds — volatility, momentum gap, model confidence level — and the system only notifies when these conditions are met. Decisions based on data, not the emotion of the moment.
Tiers are distinguished primarily by the number of concurrent API flows and the depth of backtesting available, not by hidden features.
For individual use on a limited number of monitored assets.
For traders active on several asset classes in parallel.
For desks and teams requiring dedicated API throughput.
The average latency between receiving market data and generating a signal is between 80 and 150 milliseconds, depending on the network load and the pair in question. This latency includes the processing of the model, not just the transmission of the raw stream.
Current coverage covers over 500 pairs, including major and minor currencies, major liquid cryptocurrencies and a selection of indices. The exact list depends on the level subscribed and evolves according to the liquidity available on each instrument.
Exchanges between the terminal and the infrastructure are encrypted in transit via TLS, and stored data is encrypted at rest. Access to API keys is via individual authentication, with revocation possible at any time from your account.
Join traders who already integrate Lueur Fondaru analysis into their daily preparation, before exposing their capital to the market.