fix(signals): enforce minimum 1:1 R:R on rescaled bands

fix(signals): strip <think> blocks before parsing, soften fallback copy

fix(signals): clean structured reasoning on synthesized fallback

feat(prefs): risk tolerance + trading style now shape the signal

fix(signals): tighten rescaled stop clamp to 8 percent

fix(signals): anchor entry to live price, rescale stop/target

Also replace stop/target when they are >20 percent from entry or degenerate

Re-ground parsed signals when entry drifts from live price; back-fill HOLD N/A

Fallback: if Strategist output lacks a Current Price row, call get_price_change directly

Deterministic fallback: synthesize TradingSignal from tool outputs

Simplify Strategist prompt: concrete example instead of rule-chain

Add deterministic sanity-fix for Strategist's runaway prices

Tighten Strategist prompt with explicit BUY/SELL/HOLD price inequalities

Bump to Qwen3-14B for better tool-output grounding

Swap ChatOpenAI for crewai.LLM so newer crewai versions accept the llm

Deploy FinAgent: multi-agent trading signals powered by Qwen on AMD MI300X