feat: add Anthropic Claude support and implement valuation safety che…

…cks across agents

.env.example

@@ -1,5 +1,13 @@
+# openai or anthropic
+LLM_PROVIDER=openai
+
 # Get your OpenAI API key from https://platform.openai.com/
-OPENAI_API_KEY=your-openai-api-key
+OPENAI_API_KEY=your-openai-key-here
+OPENAI_MODEL=gpt-4  # or gpt-4-turbo-preview, gpt-3.5-turbo, etc.
+
+# Get your Anthropic API key from https://console.anthropic.com/settings/keys
+ANTHROPIC_API_KEY=your-anthropic-key-here
+ANTHROPIC_MODEL=claude-3-sonnet-20240307  # or claude-3-opus-20240229, etc.
 
 # Get your Financial Datasets API key from https://financialdatasets.ai/
-FINANCIAL_DATASETS_API_KEY=your-financial-datasets-api-key
+FINANCIAL_DATASETS_API_KEY=your-financial-datasets-api-key

pyproject.toml

@@ -20,6 +20,7 @@ tabulate = "^0.9.0"
 colorama = "^0.4.6"
 questionary = "^2.1.0"
 rich = "^13.9.4"
+langchain-anthropic = "^0.3.4"
 
 [tool.poetry.group.dev.dependencies]
 pytest = "^7.4.0"

src/agents/portfolio_manager.py

@@ -2,6 +2,8 @@
 from langchain_core.messages import HumanMessage
 from langchain_core.prompts import ChatPromptTemplate
 from langchain_openai.chat_models import ChatOpenAI
+from langchain_anthropic import ChatAnthropic
+import os
 
 from graph.state import AgentState, show_agent_reasoning
 from pydantic import BaseModel, Field
@@ -156,10 +158,26 @@ def portfolio_management_agent(state: AgentState):
 
 def make_decision(prompt, tickers):
     """Attempts to get a decision from the LLM with retry logic"""
-    llm = ChatOpenAI(model="gpt-4o").with_structured_output(
-        PortfolioManagerOutput,
-        method="function_calling",
-    )
+    # Allow configurable model choice
+    model_provider = os.getenv("LLM_PROVIDER", "openai").lower()
+
+    if model_provider == "anthropic":
+        model_name = os.getenv("ANTHROPIC_MODEL", "claude-3-sonnet-20240307")
+        llm = ChatAnthropic(
+            model=model_name
+        ).with_structured_output(
+            PortfolioManagerOutput,
+            method="function_calling",
+        )
+    else:  # default to OpenAI
+        model_name = os.getenv("OPENAI_MODEL", "gpt-4")
+        llm = ChatOpenAI(
+            model=model_name
+        ).with_structured_output(
+            PortfolioManagerOutput,
+            method="function_calling",
+        )
+
     max_retries = 3
     for attempt in range(max_retries):
         try:

src/agents/valuation.py

@@ -59,8 +59,13 @@ def valuation_agent(state: AgentState):
         previous_financial_line_item = financial_line_items[1]
 
         progress.update_status("valuation_agent", ticker, "Calculating owner earnings")
-        # Calculate working capital change
-        working_capital_change = current_financial_line_item.working_capital - previous_financial_line_item.working_capital
+        # Calculate working capital change with safety checks
+        working_capital_change = 0  # Default to 0 if we can't calculate
+        if (hasattr(current_financial_line_item, 'working_capital') and 
+            hasattr(previous_financial_line_item, 'working_capital') and
+            current_financial_line_item.working_capital is not None and 
+            previous_financial_line_item.working_capital is not None):
+            working_capital_change = current_financial_line_item.working_capital - previous_financial_line_item.working_capital
 
         # Owner Earnings Valuation (Buffett Method)
         owner_earnings_value = calculate_owner_earnings_value(
@@ -74,14 +79,18 @@ def valuation_agent(state: AgentState):
         )
 
         progress.update_status("valuation_agent", ticker, "Calculating DCF value")
-        # DCF Valuation
-        dcf_value = calculate_intrinsic_value(
-            free_cash_flow=current_financial_line_item.free_cash_flow,
-            growth_rate=metrics.earnings_growth,
-            discount_rate=0.10,
-            terminal_growth_rate=0.03,
-            num_years=5,
-        )
+        # DCF Valuation with safety check for free cash flow
+        if (hasattr(current_financial_line_item, 'free_cash_flow') and 
+            current_financial_line_item.free_cash_flow is not None):
+            dcf_value = calculate_intrinsic_value(
+                free_cash_flow=current_financial_line_item.free_cash_flow,
+                growth_rate=metrics.earnings_growth,
+                discount_rate=0.10,
+                terminal_growth_rate=0.03,
+                num_years=5,
+            )
+        else:
+            dcf_value = 0
 
         progress.update_status("valuation_agent", ticker, "Comparing to market value")
         # Get the market cap
@@ -208,6 +217,11 @@ def calculate_intrinsic_value(
     Computes the discounted cash flow (DCF) for a given company based on the current free cash flow.
     Use this function to calculate the intrinsic value of a stock.
     """
+    # Add safety check for None or invalid values
+    if free_cash_flow is None or not isinstance(free_cash_flow, (int, float)) or free_cash_flow <= 0:
+        return 0
+
+
     # Estimate the future cash flows based on the growth rate
     cash_flows = [free_cash_flow * (1 + growth_rate) ** i for i in range(num_years)]
 

src/agents/warren_buffett.py

@@ -11,6 +11,8 @@
 import json
 from typing_extensions import Literal
 from utils.progress import progress
+from langchain_anthropic import ChatAnthropic
+import os
 
 
 class BuffettSignal(BaseModel):
@@ -251,12 +253,12 @@ def calculate_owner_earnings(financial_line_items: list) -> dict[str, any]:
 def calculate_intrinsic_value(financial_line_items: list) -> dict[str, any]:
     """Calculate intrinsic value using DCF with owner earnings."""
     if not financial_line_items:
-        return {"value": None, "details": ["Insufficient data for valuation"]}
+        return {"intrinsic_value": None, "details": ["Insufficient data for valuation"]}
 
     # Calculate owner earnings
     earnings_data = calculate_owner_earnings(financial_line_items)
     if not earnings_data["owner_earnings"]:
-        return {"value": None, "details": earnings_data["details"]}
+        return {"intrinsic_value": None, "details": earnings_data["details"]}
 
     owner_earnings = earnings_data["owner_earnings"]
 
@@ -265,7 +267,7 @@ def calculate_intrinsic_value(financial_line_items: list) -> dict[str, any]:
     shares_outstanding = latest_financial_line_items.outstanding_shares
 
     if not shares_outstanding:
-        return {"value": None, "details": ["Missing shares outstanding data"]}
+        return {"intrinsic_value": None, "details": ["Missing shares outstanding data"]}
 
     # Buffett's DCF assumptions
     growth_rate = 0.05  # Conservative 5% growth
@@ -340,11 +342,25 @@ def generate_buffett_output(ticker: str, analysis_data: dict[str, any]) -> Buffe
         "ticker": ticker
     })
 
-
-    llm = ChatOpenAI(model="gpt-4o").with_structured_output(
-        BuffettSignal,
-        method="function_calling",
-    )
+    # Allow configurable model choice
+    model_provider = os.getenv("LLM_PROVIDER", "openai").lower()
+
+    if model_provider == "anthropic":
+        model_name = os.getenv("ANTHROPIC_MODEL", "claude-3-sonnet-20240307")
+        llm = ChatAnthropic(
+            model=model_name
+        ).with_structured_output(
+            BuffettSignal,
+            method="function_calling",
+        )
+    else:  # default to OpenAI
+        model_name = os.getenv("OPENAI_MODEL", "gpt-4")
+        llm = ChatOpenAI(
+            model=model_name
+        ).with_structured_output(
+            BuffettSignal,
+            method="function_calling",
+        )
 
     max_retries = 3
     for attempt in range(max_retries):