Round 1: import upstream additions (TauricResearch/TradingAgents)
Pure file additions from upstream — zero conflict, zero risk to local
customizations. Cherry-picked individually rather than via merge.
New files:
- tradingagents/llm_clients/model_catalog.py — centralized model registry
- tradingagents/agents/managers/portfolio_manager.py — new PM (paired with
upstream's risk_manager.py removal, deferred to a later round)
- tradingagents/__init__.py — package init
- tests/test_google_api_key.py
- tests/test_model_validation.py
- tests/test_ticker_symbol_handling.py
- .env.enterprise.example — reference template
Skipped (deferred to later rounds):
- Round 2: bug-fix cherry-picks (yfinance retry, look-ahead bias fix,
PM-reads-trader fix, indicator normalization, etc.)
- Round 3: feat:add DeepSeek/Qwen/GLM provider support (b0f6058) —
conflicts with local llm_clients/ customizations, needs careful merge
- Round 4: portfolio manager restructure (b8b2825/318adda) — replaces
risk_manager.py which has local customizations, needs hand-port
Skipped entirely:
- Docker support (anti-Docker stack)
- Multi-language output (English-only arena)
- GPT-5.4 default flip (would undo a913416 glm-5.1 default)
- azure_client.py (not in arena roster)
- OpenRouter dynamic model selection (not in provider mix)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
7c8d523680