The system allows training agents to predict the changes of market prices of the chosen assets.

​

The data for prediction is sourced from API’s providing real time and historical data.

​

The agents are also equipped with other tools – they can source news parsed into our graph-vector DB using vector search, with data linked to specific hour-price-technical indicator nodes. They can also perform additional web searches at their discretion if deemed necessary for better forecasting... 

​

The user has the graphics UI. It allows him to see the key news represented as a thematic landscape we create using vector embeddings and annotation of cluster of news using compact large language model.