Agentic Trading

Rez's core intelligence resides in its ability to process qualitative reasoning alongside quantitative data. Unlike traditional bots that follow "If A then B" logic, Rez asks, "Why should I execute B given the current state of A?"

The Neuro-Symbolic Approach

Rez utilizes a hybrid AI architecture:

• Symbolic Layer (The Tools): Hard-coded mathematical functions (indicators like RSI, EMA) provide objective truths about market structure.

• Neural Layer (The Brain): Large Language Models (LLMs) interpret these truths. The LLM doesn't just see "RSI is 70"; it understands "RSI is 70, suggesting overbought conditions, but EMA is sloping upwards, indicating strong momentum."

The ReAct (Reasoning + Acting) Loop

Rez doesn't make decisions in a single shot. It follows a multi-turn interaction pattern known as ReAct:

1. Observation: Rez gathers the initial market context (Price, Account Balance, basic indicators).

2. Reasoning: The LLM writes out a "thought" process analyzing the data.

3. Action: If the LLM needs more data (e.g., "I want to see the 4h ATR to judge volatility"), it emits a Tool Call.

4. Observation Update: The backend executes the tool locally and feeds the results back to the LLM.

5. Final Decision: This loop repeats until the LLM has enough information to issue a buy, sell, or hold order.

Why this matters?

This approach ensures that Rez's trades are context-aware. It can decide to stay out of a "perfect" technical setup if the overall reasoning suggests high risk or upcoming volatility.

Chain of Thought Transparency

Every cycle, Rez generates a reasoning field. This is not just a log; it is the actual internal "inner monologue" the model used to arrive at its decision. This transparency allows users to understand the "why" behind every "what".