Integration

The integration between Rez's Python backend and its Next.js frontend is designed for real-time visibility and auditability. The frontend acts as a monitoring dashboard, consuming data structured by the backend.

Data Exchange Mechanism

Rez uses a file-based streaming architecture for its primary logs, ensuring that data is persisted even if the frontend or storage layer restarts.

The Trade Diary (diary.jsonl)

Every important event in the trading lifecycle is appended to the diary.jsonl file. This includes:

• Account balance snapshots.

• The agent's verbose reasoning and summary.

• Asset execution details (side, size, leverage).

• Trigger order updates (TP/SL).

The frontend reads this JSON-Lines file to populate the timeline and trade history.

Real-Time Syncing API

A lightweight aiohttp server runs alongside the trading loop. This API exposes endpoints that the frontend polls or connects to:

• /diary: Streams the latest entries from the diary.

• /state: Provides the real-time status of the trading loop (Active/Idle/Paused).

• /meta: Exposes configuration metadata (Current Risk Profile, tracked assets).

Workflow: From Trade to Dashboard

1. Backend Cycle: Rez completes a trading cycle.

2. Persistence: The decide_trade result and execution status are written to diary.jsonl.

3. API Fetch: The Next.js frontend (via SWR or React Query) fetches the latest data from the local backend API.

4. Visualization:

   • Reasoning: The "inner monologue" of the LLM is displayed in a dedicated log viewer.

   • PnL Tracker: Account value updates are plotted on a 3D-enhanced chart.

   • Active Positions: Current exposure is visualized with real-time PnL updates.

Technical Requirements for Integration

• CORS Configuration: The backend explicitly allows requests from the frontend origin (typically localhost:3000).

• JSON Schema Consistency: Both the Python backend (src/agent/) and TypeScript frontend share the same decision schema to ensure seamless parsing of trading data.