Trading agents logic

.claude/settings.local.json

@@ -0,0 +1,13 @@
+{
+  "permissions": {
+    "allow": [
+      "Bash(cd /c/Users/ethan/Desktop/ByteTrade_Repos/fintool/trading_agents && python simulate.py btc --fast 2>&1 | head -100)",
+      "Bash(python simulate.py btc --fast 2>&1 | head -80)",
+      "Bash(python simulate.py btc --fast 2>&1 | head -120)",
+      "Bash(python simulate.py btc --fast 2>&1)",
+      "Bash(pip install:*)",
+      "Bash(python simulate.py eth --fast 2>&1 && python simulate.py hype --fast 2>&1)",
+      "Bash(python simulate.py all --fast 2>&1 | tail -30)"
+    ]
+  }
+}

examples/trading_agents/README.md

@@ -0,0 +1,78 @@
+# Trading_agents
+
+The architecture of `trading_agents.py` follows `TradingAgents.md`.
+
+### Phase Hierarchy
+
+| Phase | Designation | Agents | Model | Notes |
+| :--- | :--- | :--- | :--- | :--- |
+| **Phase I** | **Eyes** | Fundamental, Technical, Sentiment, Macro | `claude-haiku-4-5` | Fast, structured data-to-text |
+| **Phase II** | **Brain** | Bull, Bear, (Facilitator loop) | `claude-opus-4-6` + Adaptive Thinking | Dialectical debate, N rounds |
+| **Phase III** | **Hands** | Trader, Risk ×3, Fund Manager | `claude-opus-4-6` + Adaptive Thinking | Weighted vote, self-reflection |
+
+---
+
+### Key Design Decisions
+
+* **Global State Management**: State is maintained as a Python `dataclass`, JSON-serialized as context for each agent to prevent hallucination drift.
+* **Reasoning Framework**: ReAct is embedded in every system prompt ("reason before concluding").
+* **Stress Testing**: Facilitator's mandatory question (*"What makes this bullish signal a trap?"*) is hardcoded into the Bear prompt for each round.
+* **Execution Safety**: 
+    * Fund Manager self-reflection is mandatory in the system prompt.
+    * A `no_trade_reason` field surfaces automatically when self-reflection triggers a halt.
+    * Confidence threshold set at **0.35**; any value below this defaults to "No Trade" regardless of suggested action.
+* **Model Optimization**: `Haiku 4.5` for Phase I (high speed/no thinking); `Opus 4.6` + Adaptive Thinking for high-stakes reasoning in Phases II/III.
+
+---
+
+### Usage
+
+**Environment Setup:**
+```bash
+export ANTHROPIC_API_KEY=sk-...
+pip install anthropic yfinance pandas
+```
+
+**Dry Run (Default):**
+```bash
+python trading_agents.py BTC --exchange hyperliquid --portfolio 1000
+```
+
+**Full Pipeline & Trade Submission:**
+```bash
+python trading_agents.py ETH --exchange binance --market perp --portfolio 5000 --rounds 3 --execute
+```
+
+**Save State:**
+```bash
+python trading_agents.py HYPE --exchange hyperliquid --output state.json
+```
+
+---
+
+### Run Simulation
+
+`cd trading_agents`
+
+#### Single scenario
+```bash
+python simulate.py btc # BTC/Hyperliquid spot → BUY
+
+python simulate.py eth # ETH/Binance perp → HOLD
+
+python simulate.py hype # HYPE/Hyperliquid spot → SELL
+```
+
+#### All three in sequence
+```bash
+python simulate.py all
+```
+
+#### Skip streaming delays (instant output)
+```bash
+python simulate.py btc --fast
+
+python simulate.py all --fast
+```
+
+No API keys or network access required — everything runs on mock data.